# Escaping the Cognitive Well: Efficient Competition Math with Off-the-Shelf Models

arXiv: https://arxiv.org/abs/2602.16793 · HTML: https://mlsanity.com/paper/2602.16793/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Placed in Mathematical reasoning, one of 17 directions in LLM reasoning.
- Builds on 10 of its 19 references that are in the map.
- First posted on arXiv 18 Feb 2026.

## Placed in

- LLM reasoning / [Mathematical reasoning](https://mlsanity.com/lr/mathematical-reasoning/)

## Builds on (most cited first, 10 of 10)

- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models](https://mlsanity.com/paper/2512.02556/) (2025)
- [MathArena: Evaluating LLMs on Uncontaminated Math Competitions](https://mlsanity.com/paper/2505.23281/) (2025)
- [DeepSeek-Prover-V2: Advancing Formal Mathematical Reasoning via Reinforcement Learning for Subgoal Decomposition](https://mlsanity.com/paper/2504.21801/) (2025)
- [Goedel-Prover-V2: Scaling Formal Theorem Proving with Scaffolded Data Synthesis and Self-Correction](https://mlsanity.com/paper/2508.03613/) (2025)
- [Goedel-Prover: A Frontier Model for Open-Source Automated Theorem Proving](https://mlsanity.com/paper/2502.07640/) (2025)
- [DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning](https://mlsanity.com/paper/2511.22570/) (2025)
- [Winning Gold at IMO 2025 with a Model-Agnostic Verification-and-Refinement Pipeline](https://mlsanity.com/paper/2507.15855/) (2025)
- [Towards Robust Mathematical Reasoning](https://mlsanity.com/paper/2511.01846/) (2025)
- [Rethinking Thinking Tokens: LLMs as Improvement Operators](https://mlsanity.com/paper/2510.01123/) (2025)

## Built on by (most cited first, 9 of 9)

- [AI co-mathematician: Accelerating mathematicians with agentic AI](https://mlsanity.com/paper/2605.06651/) (2026)
- [RMA: an Agentic System for Research-Level Mathematical Problems](https://mlsanity.com/paper/2605.22875/) (2026)
- [Pseudo-Formalization for Automatic Proof Verification](https://mlsanity.com/paper/2605.20531/) (2026)
- [OpenDeepThink: Parallel Reasoning via Bradley-Terry Aggregation](https://mlsanity.com/paper/2605.15177/) (2026)
- [AI Alignment via Incentives and Correction](https://mlsanity.com/paper/2605.01643/) (2026)
- [Self-Trained Verification for Training- and Test-Time Self-Improvement](https://mlsanity.com/paper/2605.30290/) (2026)
- [Provable Benefits of RLVR over SFT for Reasoning Models: Learning to Backtrack Efficiently](https://mlsanity.com/paper/2606.22938/) (2026)
- [Stellar Colosseum: A Many-Agent Harness for Long-Horizon Research in Mathematics and Theoretical Computer Science](https://mlsanity.com/paper/2609.15983/) (2026)
- [Splicing Hints into Reasoning Traces: Learning from Demonstrator Failures](https://mlsanity.com/paper/2610.13996/) (2026)

## Abstract (arXiv)

In the past year, custom and unreleased math reasoning models reached gold medal performance on the International Mathematical Olympiad (IMO). Similar performance was then reported using large-scale inference on publicly available models but at prohibitive costs (e.g., 3000 USD per problem). In this work, we present an inference pipeline that attains best-in-class performance on IMO-style math problems at an average inference cost orders of magnitude below competing methods while using only general-purpose off-the-shelf models. Our method relies on insights about grader failure in solver-grader pipelines, which we call the Cognitive Well (iterative refinement converging to a wrong solution that the solver as well as the pipeline's internal grader consider to be basically correct). Our pipeline addresses these failure modes through conjecture extraction, wherein candidate lemmas are isolated from generated solutions and independently verified alongside their negations in a fresh environment (context detachment). On IMO-ProofBench Advanced (PB-Adv), our pipeline achieves 67.1 percent performance using Gemini 3.0 Pro with an average cost per question of approximately 31 USD. At the time of evaluation, this represented the state-of-the-art on PB-Adv among both public and unreleased models, and more than doubles the success rate of the next best publicly accessible pipeline, all at a fraction of the cost.