Test-time Recursive Thinking: Self-Improvement without External Feedback
This addresses the problem of reducing reliance on external training data for LLM improvement, though it appears incremental as it builds on existing self-improvement concepts.
The paper tackles the problem of enabling Large Language Models (LLMs) to self-improve without external feedback by addressing challenges in generating diverse solutions and selecting correct answers, resulting in 100% accuracy on AIME-25/24 and 10.4-14.8 percentage point improvements on LiveCodeBench's hardest problems.
Modern Large Language Models (LLMs) have shown rapid improvements in reasoning capabilities, driven largely by reinforcement learning (RL) with verifiable rewards. Here, we ask whether these LLMs can self-improve without the need for additional training. We identify two core challenges for such systems: (i) efficiently generating diverse, high-quality candidate solutions, and (ii) reliably selecting correct answers in the absence of ground-truth supervision. To address these challenges, we propose Test-time Recursive Thinking (TRT), an iterative self-improvement framework that conditions generation on rollout-specific strategies, accumulated knowledge, and self-generated verification signals. Using TRT, open-source models reach 100% accuracy on AIME-25/24, and on LiveCodeBench's most difficult problems, closed-source models improve by 10.4-14.8 percentage points without external feedback.