task stringclasses 1
value | input stringlengths 259 305 | output stringclasses 5
values | type stringclasses 1
value | focus stringclasses 1
value | difficulty stringclasses 1
value |
|---|---|---|---|---|---|
coding | Coding Hard 184: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1603: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1539: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 532: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1733: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1647: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1325: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 893: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1436: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 264: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1931: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 261: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 298: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1794: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1152: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 752: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1847: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 668: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1445: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 939: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 947: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 658: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1990: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1703: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1908: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 810: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1005: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1555: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 750: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1680: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 568: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1421: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 819: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1017: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1163: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1428: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1495: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1656: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1125: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 644: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1800: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 966: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 781: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1021: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 824: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1281: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 97: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 52: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 364: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1777: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1808: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 637: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1508: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1661: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1198: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1583: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 634: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1690: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1423: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1962: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1237: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 411: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 423: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1168: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1838: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 614: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1371: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 452: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 913: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 474: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1463: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 121: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1585: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 443: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1867: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 346: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1006: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1334: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 551: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 704: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1738: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1886: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 80: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1456: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 204: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 951: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 501: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1701: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1574: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1242: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1550: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1130: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 849: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 712: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1899: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1189: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 106: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1174: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 510: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 419: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.