AI Paper Atlas

Research map / Large language models / Agents, code and math

Unleashing Scientific Reasoning for Bio-experimental Protocol Generation via Structured Component-based Reward Mechanism

Haoran Sun, Yankai Jiang, Zhenyu Tang, Yaning Pan, Shuang Gu, Zekai Lin, Lilong Wang, Wenjie Lou, Lei Liu, LEI BAI, Xiaosong Wang

ICLR 2026 · Large language models › Agents, code and math · approach: LLMs for scientific discovery · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add step-count reward f(d) (ablation)Thoth · SciRecipe-EvalStep-M: 10.00 -> 53.00Improved
add length penalty g(L) (ablation)Thoth · SciRecipe-EvalSemantic-A: 38.89 -> 46.60Improved
add positional decay factor m_ij (ablation)Thoth · SciRecipe-EvalSemantic-A: 38.68 -> 46.60Improved
add Order consistency reward (ablation)Thoth · SciRecipe-EvalOrder-S: 12.83 -> 25.50Improved
add KL penalty (ablation)Thoth · SciRecipe-EvalAVG: 48.89 -> 52.10Improved
SCORE reward instead of vanilla lexical reward (ablation)Thoth · SciRecipe-EvalAVG: 45.62 -> 52.10Improved
SCORE reward instead of vanilla lexical reward (ablation)Thoth · SciRecipe-EvalBLEU-AVG: 45.52 -> 43.62Hurt
add scientific review (SciCheck) data (ablation)Thoth · SciRecipe-EvalAVG: 47.37 -> 49.45Improved
add Sketch-and-Fill paradigm (ablation)Qwen3-8B · SciRecipe-EvalBLEU-AVGNo clear effect
add Sketch-and-Fill paradigm (ablation)DeepSeek-V3, GPT-5 Chat, Thoth · SciRecipe-Evallexical similarity avg: +3.89%, +2.92%, +3.79%Improved
three-stage training (pretrain+SFT+RL) vs Stage 1+2 (ablation)Thoth · SciRecipe-Evalexecutability avg: +11.08%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy