# Reasoning or Retrieval? A Study of Answer Attribution on Large Reasoning Models

arXiv: https://arxiv.org/abs/2509.24156 · HTML: https://mlsanity.com/paper/2509.24156/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 43 of its 61 references that are in the map.
- First posted on arXiv 29 Sep 2025.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Builds on (most cited first, 30 of 43)

- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [Tree of Thoughts: Deliberate Problem Solving with Large Language Models](https://mlsanity.com/paper/2305.10601/) (2023)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [s1: Simple test-time scaling](https://mlsanity.com/paper/2501.19393/) (2025)
- [Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting](https://mlsanity.com/paper/2305.04388/) (2023)
- [Tulu 3: Pushing Frontiers in Open Language Model Post-Training](https://mlsanity.com/paper/2411.15124/) (2024)
- [Mass-Editing Memory in a Transformer](https://mlsanity.com/paper/2210.07229/) (2022)
- [Group Sequence Policy Optimization](https://mlsanity.com/paper/2507.18071/) (2025)
- [Phi-4 Technical Report](https://mlsanity.com/paper/2412.08905/) (2024)
- [Measuring Faithfulness in Chain-of-Thought Reasoning](https://mlsanity.com/paper/2307.13702/) (2023)
- [SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training](https://mlsanity.com/paper/2501.17161/) (2025)
- [Dissecting Recall of Factual Associations in Auto-Regressive Language Models](https://mlsanity.com/paper/2304.14767/) (2023)
- [Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning](https://mlsanity.com/paper/2404.05868/) (2024)
- [Reasoning Models Don't Always Say What They Think](https://mlsanity.com/paper/2505.05410/) (2025)
- [Chain-of-Thought Reasoning In The Wild Is Not Always Faithful](https://mlsanity.com/paper/2503.08679/) (2025)
- [H-CoT: Hijacking the Chain-of-Thought Safety Reasoning Mechanism to Jailbreak Large Reasoning Models, Including OpenAI o1/o3, DeepSeek-R1, and Gemini 2.0 Flash Thinking](https://mlsanity.com/paper/2502.12893/) (2025)
- [Are DeepSeek R1 And Other Reasoning Models More Faithful?](https://mlsanity.com/paper/2501.08156/) (2025)
- [The Benefits of a Concise Chain of Thought on Problem-Solving in Large Language Models](https://mlsanity.com/paper/2401.05618/) (2024)
- [Frequency-Aligned Knowledge Distillation for Lightweight Spatiotemporal Forecasting](https://mlsanity.com/paper/2507.02939/) (2025)
- [Characterizing Mechanisms for Factual Recall in Language Models](https://mlsanity.com/paper/2310.15910/) (2023)
- [On the Hardness of Faithful Chain-of-Thought Reasoning in Large Language Models](https://mlsanity.com/paper/2406.10625/) (2024)
- [Competition of Mechanisms: Tracing How Language Models Handle Facts and Counterfactuals](https://mlsanity.com/paper/2402.11655/) (2024)
- [Towards Robust and Secure Embodied AI: A Survey on Vulnerabilities and Attacks](https://mlsanity.com/paper/2502.13175/) (2025)

## Built on by (most cited first, 10 of 10)

- [AgentLAB: Benchmarking LLM Agents against Long-Horizon Attacks](https://mlsanity.com/paper/2602.16901/) (2026)
- [GaussianDWM: 3D Gaussian Driving World Model for Unified Scene Understanding and Multi-Modal Generation](https://mlsanity.com/paper/2512.23180/) (2025)
- [RAM: Recover Any 3D Human Motion in-the-Wild](https://mlsanity.com/paper/2603.19929/) (2026)
- [RASA: Routing-Aware Safety Alignment for Mixture-of-Experts Models](https://mlsanity.com/paper/2602.04448/) (2026)
- [Measuring and Mitigating Post-hoc Rationalization in Reverse Chain-of-Thought Generation](https://mlsanity.com/paper/2602.14469/) (2026)
- [Measuring and curing reasoning rigidity: from decorative chain-of-thought to genuine faithfulness](https://mlsanity.com/paper/2603.22816/) (2026)
- [Hypergraph and Latent ODE Learning for Multimodal Root Cause Localization in Microservices](https://mlsanity.com/paper/2605.00351/) (2026)
- [Reasoning or Memorization: Can LLMs Understand and Generate Chinese Xiehouyu Riddles?](https://mlsanity.com/paper/2607.23440/) (2026)
- [MARS: Multi-Agent Re-ranking for Repeat-Order Food Delivery Recommendation](https://mlsanity.com/paper/2607.25420/) (2026)
- [BrainTRACE: Tracing Longitudinal, Multimodal, and Volumetric Evidence in Brain MRI Clinical Reasoning](https://mlsanity.com/paper/2610.06571/) (2026)

## Abstract (arXiv)

Large reasoning models (LRMs) exhibit unprecedented capabilities in solving complex problems through Chain-of-Thought (CoT) reasoning. However, recent studies reveal that their final answers often contradict their own reasoning traces. We hypothesize that this inconsistency stems from two competing mechanisms for generating answers: CoT reasoning and memory retrieval. To test this hypothesis, we conduct controlled experiments that challenge LRMs with misleading cues during reasoning and/or corrupted answers during retrieval. Our results across models and datasets confirm that both mechanisms operate simultaneously, with their relative dominance influenced by multiple factors: problem domains, model scales, and fine-tuning approaches (e.g., reinforcement learning vs. distillation). The findings reveal a critical limitation in current reasoning fine-tuning paradigms: models can exploit the retrieval mechanism as a shortcut, effectively "hacking" the reward signal and undermining genuine reasoning development. To address this challenge, we introduce FARL, a novel fine-tuning framework that integrates memory unlearning with reinforcement learning. By carefully suppressing retrieval shortcuts during the fine-tuning process, FARL promotes reasoning-dominant behavior and enhances generalizable reasoning capabilities. The code is available: https://github.com/ZJUWYH/FARL.