heldout lab

heldoutlab's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in prism-ml/Ternary-Bonsai-2-27B-gguf about 8 hours ago

Independent test: 98% retention on easier tasks, 87% on harder reasoning/coding (vs 4-bit Qwen3.8 27B)

🔥❤️ 7

17

#102 opened 19 days ago by

heldoutlab

New activity in LiquidAI/LFM2.5-24B-A2B 5 days ago

Independent tool-calling test vs Qwen3.6-35B-A3B (Q4_K_M, llama.cpp)

3

#12 opened 16 days ago by

heldoutlab

New activity in unsloth/gemma-4-26B-A4B-it-GGUF 7 days ago

q8_0 KV cache test: decode up to 47% slower on Apple Silicon, accuracy not resolved

1

#47 opened 7 days ago by

heldoutlab

New activity in LiquidAI/LFM2.5-24B-A2B-GGUF 16 days ago

Q4_K_M on llama.cpp: tool calling works, but 49% vs Qwen3.6's 91% on 75 multi-step tasks

➕ 1

#9 opened 16 days ago by

heldoutlab

New activity in prism-ml/Ternary-Bonsai-2-27B-gguf 17 days ago

Has anyone compared Ternary Bonsai 2 model quality retention versus ISTA-DASLab's GSQ-RCO?

👍 1

11

#99 opened 24 days ago by

ssvb

New activity in prism-ml/Ternary-Bonsai-2-27B-gguf 18 days ago

Independent test: 98% retention on easier tasks, 87% on harder reasoning/coding (vs 4-bit Qwen3.8 27B)

🔥❤️ 7

17

#102 opened 19 days ago by

heldoutlab

Has anyone compared Ternary Bonsai 2 model quality retention versus ISTA-DASLab's GSQ-RCO?

👍 1

11

#99 opened 24 days ago by

ssvb

Has anyone compared Ternary Bonsai 2 model quality retention versus ISTA-DASLab's GSQ-RCO?

👍 1

11

#99 opened 24 days ago by

ssvb

Independent test: 98% retention on easier tasks, 87% on harder reasoning/coding (vs 4-bit Qwen3.8 27B)

🔥❤️ 7

17

#102 opened 19 days ago by

heldoutlab