Rethinking the Unsolvable: When In-Context Search Meets Test-Time Scaling

TL;DR

This paper demonstrates that combining in-context search prompting with test-time scaling significantly enhances LLM reasoning performance on complex, previously unsolvable problems, challenging existing evaluation paradigms.

Contribution

It introduces a novel approach that leverages advanced in-context search and internal scaling to substantially improve LLM reasoning capabilities on hard tasks.

Findings

Achieves up to 30x success rate improvement on complex tasks

Empirically validates the approach on NP-hard and real-world benchmarks

Theoretically extends the class of solvable reasoning problems

Abstract

Recent research has highlighted that Large Language Models (LLMs), even when trained to generate extended long reasoning steps, still face significant challenges on hard reasoning problems. However, much of the existing literature relies on direct prompting with simple in-context learning examples for evaluation, which largely overlooks advanced techniques to elicit LLMs' deliberate reasoning before drawing conclusions that LLMs hit a performance ceiling. In this paper, we systematically explore the combined potential of in-context search and test-time scaling on super hard reasoning tasks. We find that by employing advanced in-context search prompting to LLMs augmented with internal scaling, one can achieve transformative performance breakthroughs on tasks previously deemed "unsolvable" (e.g., reported success rates below 5%). We provide both empirical results and theoretical analysis of how this combination can unleash LLM reasoning capabilities: i) Empirically, on controlled NP-hard tasks and complex real-world planning benchmarks, our approach achieves up to a 30x improvement in success rates compared to previously reported results without any external mechanisms; ii) Theoretically, we show that in-context search prompting, when combined with internal scaling, significantly extends the complexity class of solvable reasoning problems. These findings challenge prevailing assumptions about the limitations of LLMs on complex tasks, indicating that current evaluation paradigms systematically underestimate their true potential. Our work calls for a critical reassessment of how LLM reasoning is benchmarked and a more robust evaluation strategy that fully captures the true capabilities of contemporary LLMs, which can lead to a better understanding of their operational reasoning boundaries in real-world deployments.