Paper Detail

Beyond Correctness: Benchmarking and Aligning Response Behaviors in Hybrid-Thinking MLLMs

Xinming Wang, Weinong Wang, Hongming Yang, Yansong Lin, Zheng Ruan, Shangpin Peng, Qiming Peng, Nan Qiao, Fengyuan Lu, Guoqing Ma, Marito Li, Songyang Zhang, Saiyong Yang, Han Hu, Yonglong Tian, Xu-Yao Zhang

huggingface Score 21.5

Published 2026-08-17 · First seen 2026-08-24

General AI

Abstract

Hybrid-thinking multimodal large language models (MLLMs) allow a single model to alternate between deliberative thinking and latency-efficient non-thinking inference. Although these modes differ in reasoning budget, their delivered responses should satisfy the same user-facing standard. Correctness alone may not characterize this response quality; we therefore evaluate task accuracy and response-pattern failures as complementary outcomes. We study this gap through response-pattern alignment: whether thinking and non-thinking interfaces preserve acceptable final-response behavior. We introduce PatternEval, a failure-enriched diagnostic benchmark comprising 2,415 multimodal prompts spanning visual perception and grounding, structured image understanding, and multimodal knowledge reasoning. PatternEval tests four recurrent failures: chain-of-thought leakage, response repetition, logical contradiction, and performative reasoning. Response-pattern failures are widespread across models from different providers, with non-thinking inference exhibiting substantially higher failure rates and thereby creating systematic misalignment between thinking and non-thinking interfaces. Motivated by this diagnosis, we develop PatternRM, a response-level reward model, and PatternRL, which introduces pattern-specific penalties during reinforcement learning. Experiments on Qwen3-VL-4B and Qwen3-VL-8B show that incorporating pattern-specific penalties into reinforcement learning can mitigate cross-mode misalignment while incurring a marginal task performance trade-off. Together, PatternEval and PatternRL provide an evaluation-and-training framework for aligning user-visible response patterns across hybrid-thinking interfaces.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
now
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@misc{wang2026beyond,
  title = {Beyond Correctness: Benchmarking and Aligning Response Behaviors in Hybrid-Thinking MLLMs},
  author = {Xinming Wang and Weinong Wang and Hongming Yang and Yansong Lin and Zheng Ruan and Shangpin Peng and Qiming Peng and Nan Qiao and Fengyuan Lu and Guoqing Ma and Marito Li and Songyang Zhang and Saiyong Yang and Han Hu and Yonglong Tian and Xu-Yao Zhang},
  year = {2026},
  abstract = {Hybrid-thinking multimodal large language models (MLLMs) allow a single model to alternate between deliberative thinking and latency-efficient non-thinking inference. Although these modes differ in reasoning budget, their delivered responses should satisfy the same user-facing standard. Correctness alone may not characterize this response quality; we therefore evaluate task accuracy and response-pattern failures as complementary outcomes. We study this gap through response-pattern alignment: whe},
  url = {https://huggingface.co/papers/2608.12781},
  keywords = {multimodal large language models, hybrid-thinking, response-pattern alignment, PatternEval, chain-of-thought leakage, logical contradiction, performative reasoning, PatternRM, PatternRL, reinforcement learning, huggingface daily},
  eprint = {2608.12781},
  archiveprefix = {arXiv},
}

Metadata

{}