Socratic-MCTS: Test-Time Visual Reasoning by Asking the Right Questions

TL;DR

This paper introduces Socratic-MCTS, a search-based method that prompts vision-language models with subquestions to enhance their reasoning capabilities without additional training, leading to improved performance on reasoning benchmarks.

Contribution

It proposes a novel MCTS-inspired algorithm that elicits reasoning in pre-trained models by injecting subquestions, enabling extended reasoning without retraining.

Findings

Achieves a 2% overall improvement on MMMU-PRO benchmark.

Yields a 9% gain in Liberal Arts category.

Demonstrates consistent reasoning improvements across three benchmarks.

Abstract

Recent research in vision-language models (VLMs) has centered around the possibility of equipping them with implicit long-form chain-of-thought reasoning -- akin to the success observed in language models -- via distillation and reinforcement learning. But what about the non-reasoning models already trained and deployed across the internet? Should we simply abandon them, or is there hope for a search mechanism that can elicit hidden knowledge and induce long reasoning traces -- without any additional training or supervision? In this paper, we explore this possibility using a Monte Carlo Tree Search (MCTS)-inspired algorithm, which injects subquestion-subanswer pairs into the model's output stream. We show that framing reasoning as a search process -- where subquestions act as latent decisions within a broader inference trajectory -- helps the model "connect the dots" between fragmented knowledge and produce extended reasoning traces in non-reasoning models. We evaluate our method across three benchmarks and observe consistent improvements. Notably, our approach yields a 2% overall improvement on MMMU-PRO, including a significant 9% gain in Liberal Arts.