Models from "Efficient Reasoning Training Does Not Always Harm CoT Faithfulness and Monitorability".
Samuel Lewis-Lim PRO
Samll
AI & ML interests
None yet
Recent Activity
updated a collection about 23 hours ago
efficient-reasoning submitted a paper about 23 hours ago
Efficient Reasoning Training Does Not Always Harm CoT Faithfulness and Monitorability upvoted a paper 4 days ago
Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical PriorOrganizations
None yet
models 11
Samll/olmo3-7b-think-l1
Updated • 12
Samll/olmo3-7b-think-glp
Updated • 11
Samll/olmo3-7b-think-thinkprune-b4000
Text Generation • 7B • Updated • 171
Samll/qwen3-8b-l1
Updated • 9
Samll/qwen3-8b-glp
Updated • 10
Samll/qwen3-8b-thinkprune-b3000
Text Generation • 8B • Updated • 213
Samll/qwen3-4b-thinking-l1
Updated • 11
Samll/qwen3-4b-thinking-glp
Updated • 8
Samll/qwen3-4b-thinking-thinkprune-b4000
Text Generation • 4B • Updated • 252
Samll/wild_hallu_v3-3b
Text Generation • 3B • Updated • 11
datasets 20
Samll/wild_hallucinations_deepseek_factscore_chat_v2
Viewer • Updated • 3.65k • 36
Samll/wild_hallucinations_deepseek_factscore_chat
Viewer • Updated • 3.38k • 30
Samll/hotpotqa_inference_llama3-8b
Viewer • Updated • 20k • 9
Samll/wild_hallucinations_deepseek_factscore
Viewer • Updated • 1k • 32
Samll/wild_hallucinations_with_responses
Viewer • Updated • 7.92k • 39
Samll/Llama-3.1-8B-Instruct_attn_rollout_attributions_v1
Viewer • Updated • 5.77k • 31
Samll/Llama-3.1-8B-Instruct_attn_rollout_attributions
Viewer • Updated • 15.1k • 24
Samll/Llama-3.1-8B-Instruct_Taur_CoT_contrastive_relevances_attn_test
Viewer • Updated • 210 • 15
Samll/Llama-3.1-8B-Instruct_Taur_CoT_contrastive_relevances_v1
Viewer • Updated • 13.3k • 40
Samll/Llama-3.1-8B-Instruct_Taur_CoT_contrastive_relevances
Viewer • Updated • 22.4k • 31