Research map / Large language models / Reasoning and chain of thought
OncoCoT: A Temporal-causal Chain-of-Thought Dataset for Oncologic Decision-Making
Peiru Yang, Yudong Li, Shiting Wang, Xinyi Liu, Haotian Gan, Xintian Li, Qingyu Gao, Yongfeng Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Introduces OncoCoT, a chain-of-thought dataset for oncologic decision-making that models temporal-causal dependencies from longitudinal clinical outcomes.
Problem
Long chain-of-thought reasoning is hard to verify in medical decision-making, where validation depends on longitudinal clinical outcomes with temporal-causal dependencies rather than static verification frameworks.
Approach
A temporal-causal chain-of-thought dataset for oncology that grounds reasoning traces in longitudinal clinical outcomes, so that reasoning steps can be validated over time rather than by static checks.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune on OncoCoT instead of no fine-tuning | R1-Distill-Qwen-7B · OncoEval Treatment | avg score: 39.62 -> 50.13 | Improved |
| fine-tune on OncoCoT instead of no fine-tuning | R1-Distill-Qwen-7B · OncoEval Examination | avg score: 28.51 -> 34.95 | Improved |
| fine-tune on OncoCoT instead of no fine-tuning | R1-Distill-Qwen-7B · OncoEval Diagnosis | avg score: 44.16 -> 44.55 | No clear effect |
| use temporal-logic constrained CoT instead of rejection-sampling CoT (ablation) | R1-Distill-Qwen-7B · OncoEval Treatment | avg score: 46.40 -> 50.13 | Improved |
| use temporal-logic constrained CoT instead of rejection-sampling CoT (ablation) | R1-Distill-Qwen-7B · OncoEval Examination | avg score: 33.16 -> 34.95 | Improved |
| add CoT reasoning to Q&A fine-tuning data (R1-distill CoT vs w/o CoT) (ablation) | R1-Distill-Qwen-7B · OncoEval Diagnosis | avg score: 41.78 -> 43.92 | Improved |
| fine-tune on Q&A without CoT (ablation) | R1-Distill-Qwen-7B · OncoEval Diagnosis | avg score: 44.16 -> 41.78 | Hurt |
| use larger or closed-source reasoning models | frontier LLMs · OncoEval | accuracy | No clear effect |
Closest papers by meaning
- MTBBench: A Multimodal Sequential Clinical Decision-Making Benchmark in Oncology (NeurIPS 2025 · similarity 0.89)
- TumorChain: Interleaved Multimodal Chain-of-Thought Reasoning for Traceable Clinical Tumor Analysis (ICLR 2026 · similarity 0.87)
- MedS³: Towards Medical Slow Thinking with Self-Evolved Soft Dual-sided Process Supervision (AAAI 2026 · similarity 0.86)
- M3CoTBench: Benchmark Chain-of-Thought of MLLMs in Medical Image Understanding (ICLR 2026 · similarity 0.85)
- Optimizing Inference-Time Compute for Medical Reasoning via Uncertainty Quantification (ICML 2026 · similarity 0.85)
- DoAtlas-1: A Causal Compilation Paradigm for Clinical AI (NeurIPS 2026 · similarity 0.84)
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.
