AI Paper Atlas

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

Pushing the Boundaries of Natural Reasoning: Interleaved Bonus from Formal-Logic Verification

Chuxue Cao, Jinluan Yang, Haoran Li, Kunhao Pan, Zijian Zhao, Zhengyu Chen, Yuchen Tian, Lijun Wu, Conghui He, Yike Guo, Sirui Han

ICML 2026 · Large language models › Agents, code and math · approach: Inductive and abstract reasoning evaluation · 0 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 formal logic verification to SFT data (FLV-SFT vs Natural-SFT) (ablation)Qwen2.5-7B SFT · six-benchmark averageaverage accuracy: 36.3 -> 49.8Improved
add formal logic verification to SFT data (ablation)Qwen2.5-7B SFT · MATH500accuracy: 56.6 -> 77.2Improved
add GRPO with composite reward after FLV-SFT (ablation)FLV-SFT Qwen2.5-7B · six-benchmark averageaverage accuracy: 49.8 -> 51.9Improved
add GRPO after FLV-SFT (ablation)FLV-SFT Qwen2.5-14B · six-benchmark averageaverage accuracy: 55.7 -> 58.6Improved
add RL to natural-language SFT model (no formal verification) (ablation)Natural-SFT Qwen2.5-7B · six-benchmark averageaverage accuracy: 36.3 -> 37.0No clear effect
skip SFT cold start, apply RL directly to base model (FLV-ZeroRL) (ablation)FLV-RL Qwen2.5-7B · six-benchmark averageaverage accuracy: 51.9 -> 42.7Hurt
replace enforced per-step verification with flexible verification (decouple calculation from validation) (ablation)FLV training data · math, logic, general subsetsaccuracyImproved
formal logic verification-guided RL vs tool-integrated SimpleTIRQwen2.5-7B · KOR-Benchaccuracy: 37.0 -> 51.0Improved
formal logic verification-guided RL vs tool-integrated SimpleTIRQwen2.5-7B · AIME24avg@16 accuracy: 41.0 -> 20.8Hurt
formal logic verification-guided RL vs General-ReasonerQwen2.5-7B · GPQA-Diamondaccuracy: 38.9 -> 35.4Hurt

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