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 1333: 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 232: 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 1073: 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 860: 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 1110: 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 731: 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 1251: 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 332: 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 1004: 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 1898: 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 1473: 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 1394: 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 1083: 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 1181: 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 329: 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 1695: 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 1239: 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 116: 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 993: 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 1426: 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 208: 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 935: 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 1923: 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 1260: 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 1577: 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 1940: 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 915: 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 1890: 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 391: 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 1952: 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 1477: 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 589: 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 759: 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 28: 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 75: 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 446: 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 1020: 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 58: 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 902: 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 1971: 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 488: 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 117: 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 1837: 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 183: 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 7: 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 1471: 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 1802: 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 1989: 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 932: 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 152: 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 133: 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 1770: 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 1053: 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 394: 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 1929: 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 321: 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 227: 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 1285: 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 1244: 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 794: 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 59: 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 162: 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 299: 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 1759: 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 807: 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 467: 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 636: 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 242: 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 1431: 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 168: 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 502: 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 1746: 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 994: 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 1610: 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 1154: 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 1941: 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 1304: 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 1678: 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 899: 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 148: 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 673: 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 1075: 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 1512: 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 1324: 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 677: 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 358: 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 335: 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 1217: 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 1724: 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 376: 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 429: 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 157: 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 1442: 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 1101: 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 1430: 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 642: 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 1763: 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 1732: 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 984: 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 1476: 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 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.