# CORE: Contrastive Reflection Enables Rapid Improvements in Reasoning

arXiv: https://arxiv.org/abs/2605.28742 · HTML: https://mlsanity.com/paper/2605.28742/

## Key facts

- Cited by 1 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 24 of its 35 references that are in the map.
- First posted on arXiv 27 May 2026.

## 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, 24 of 24)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Reflexion: Language Agents with Verbal Reinforcement Learning](https://mlsanity.com/paper/2303.11366/) (2023)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [Voyager: An Open-Ended Embodied Agent with Large Language Models](https://mlsanity.com/paper/2305.16291/) (2023)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [STaR: Bootstrapping Reasoning With Reasoning](https://mlsanity.com/paper/2203.14465/) (2022)
- [GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning](https://mlsanity.com/paper/2507.19457/) (2025)
- [The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity](https://mlsanity.com/paper/2506.06941/) (2025)
- [Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs](https://mlsanity.com/paper/2406.11695/) (2024)
- [ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memory](https://mlsanity.com/paper/2509.25140/) (2025)
- [Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety](https://mlsanity.com/paper/2507.11473/) (2025)
- [ZebraLogic: On the Scaling Limits of LLMs for Logical Reasoning](https://mlsanity.com/paper/2502.01100/) (2025)
- [MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory](https://mlsanity.com/paper/2601.03192/) (2026)
- [Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward](https://mlsanity.com/paper/2602.12430/) (2026)
- [Jina Embeddings 2: 8192-Token General-Purpose Text Embeddings for Long Documents](https://mlsanity.com/paper/2310.19923/) (2023)
- [When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors](https://mlsanity.com/paper/2507.05246/) (2025)
- [Expanding the Capabilities of Reinforcement Learning via Text Feedback](https://mlsanity.com/paper/2602.02482/) (2026)
- [Metacognitive Reuse: Turning Recurring LLM Reasoning Into Concise Behaviors](https://mlsanity.com/paper/2509.13237/) (2025)
- [Experiential Reinforcement Learning](https://mlsanity.com/paper/2602.13949/) (2026)
- [Useful Memories Become Faulty When Continuously Updated by LLMs](https://mlsanity.com/paper/2605.12978/) (2026)
- [MathGAP: Out-of-Distribution Evaluation on Problems with Arbitrarily Complex Proofs](https://mlsanity.com/paper/2410.13502/) (2024)
- [Systematic human learning and generalization from a brief tutorial with explanatory feedback](https://mlsanity.com/paper/2107.06994/) (2021)
- [Leveraging Speech to Identify Signatures of Insight and Transfer in Problem Solving](https://mlsanity.com/paper/2605.12970/) (2026)

## Built on by (most cited first, 1 of 1)

- [RLTL;DR: Self-improvement by Internalizing Self-generated Feedback](https://mlsanity.com/paper/2609.37633/) (2026)

## Abstract (arXiv)

Language models can use verifiable rewards to improve at a wide variety of reasoning tasks. However, both parametric (e.g. RLVR) and non-parametric (e.g. prompt optimization) approaches to doing so typically require hundreds of training samples and thousands of model rollouts, making them expensive in the best case and intractable in the worst. To address this challenge, we introduce Contrastive Reflection (CORE), a non-parametric learning algorithm that compares past reasoning traces to generate insights: short natural-language descriptions of reasoning strategies and constraints that capture differences between successful and unsuccessful problem attempts. Across four reasoning tasks, we demonstrate that CORE enables more rapid improvement than both parametric (GRPO) and non-parametric (GEPA, episodic RAG, and MemRL) methods, while using fewer rollouts. Under fixed rollout budgets with as few as five training samples, CORE achieves the strongest performance in most task-data regimes. Finally, we highlight how CORE is substantially more context-efficient than non-parametric baselines, requiring fewer prompt tokens while storing learned knowledge as compact, interpretable natural-language insights. Our results therefore suggest that distilling contrasts between successful and unsuccessful reasoning traces into abstract and useful insights can provide a more efficient and interpretable route to model self-improvement than weight updates, prompt optimization, or direct reuse of stored reasoning traces.