Fan Yang, Wentao Zhang, Tianpeng Li, Hao Liang, Zenan Zhou, Linzhuang Sun, Jingxuan Wei, Bihui Yu
We lifted 4 functions out of this paper's own repositories and ran 0 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| Aurora-slz/MM-Verify | — | 0 of 4 |
| Function | Status | Where it lives |
|---|---|---|
| MCTS | Not yet run | Aurora-slz/MM-Verify/search/MCTS/mcts.py pointer only (licence: NONE) · get_code("856c9aad637b0958") |
| MCTS_search | Not yet run | Aurora-slz/MM-Verify/search/MCTS/mcts.py pointer only (licence: NONE) · get_code("040a0e7f4b118e77") |
| executeRound | Not yet run | Aurora-slz/MM-Verify/search/MCTS/mcts.py pointer only (licence: NONE) · get_code("cdde1fa1001fca90") |
| expand | Not yet run | Aurora-slz/MM-Verify/search/MCTS/mcts.py pointer only (licence: NONE) · get_code("b694a3efa9770e88") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
According to the Test-Time Scaling, the integration of External Slow-Thinking with the Verify mechanism has been demonstrated to enhance multi-round reasoning in large language models (LLMs). However, in the multimodal (MM) domain, there is still a lack of a strong MM-Verifier. In this paper, we introduce MM-Verifier and MM-Reasoner to enhance multimodal reasoning through longer inference and more robust verification. First, we propose a two-step MM verification data synthesis method, which combines a simulation-based tree search with verification and uses rejection sampling to generate high-quality Chainof-Thought (COT) data. This data is then used to fine-tune the verification model, MM-Verifier. Additionally, we present a more efficient method for synthesizing MMCOT data, bridging the gap between text-based and multimodal reasoning. The synthesized data is used to fine-tune MM-Reasoner. Our MM-Verifier outperforms all larger models on the MathCheck, MathVista, and MathVerse benchmarks. Moreover, MM-Reasoner demonstrates strong effectiveness and scalability, with performance improving as data size increases. Finally, our approach achieves strong performance when combining MM-Reasoner and MM-Verifier, reaching an accuracy of 65.3 on MathVista, surpassing GPT-4o (63.8) with 12 rollouts. Our code is made available https: //github.com/Aurora-slz/MM-Verify.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2502.13383")
get_code_for_paper("2502.13383")
have("2502.13383")
Connect an agent — have() is free.