Debias in Text, Believe Your Eyes: Text-Anchored Cross-Modal Transfer for Visual Counter-Commonsense Reasoning

• URL: http://arxiv.org/abs/2608.06938v1
• Date: Fri, 07 Aug 2026 08:10:31 GMT
• Title: Debias in Text, Believe Your Eyes: Text-Anchored Cross-Modal Transfer for Visual Counter-Commonsense Reasoning
• Abstract summary: Recent studies mainly improve visual counter-commonsense reasoning by enhancing visual inputs.<n>We introduce TACT, a text-anchored post-training framework that debiases the shared language decoder without requiring any visual training data.<n>TACT substantially improves visual reasoning while preserving general capabilities, demonstrating effective text-to-vision cross-modal transfer.
• Score: 21.36422191204697
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The visual reasoning ability of multimodal large language models (MLLMs) is crucial for downstream applications, particularly counter-commonsense reasoning, which requires models to reason beyond common assumptions. Recent studies mainly improve visual counter-commonsense reasoning by enhancing visual inputs, following the assumption that failures originate from insufficient visual grounding. However, our empirical analysis reveals that the bottleneck is not visual perception. MLLMs already capture the relevant visual evidence, and the correct answer exists in their decoding space. Instead, the shared language decoder resolves prior--evidence conflicts by favoring dominant language priors, especially for low-frequency factual scenarios. Motivated by this, we first propose a text-anchored data construction pipeline, whose core component, Fact-Frequency Distillation (FFD), estimates the prior strength of commonsense facts and distills verified counter-commonsense scenarios into a high-quality text corpus. Building upon this corpus, we introduce TACT, a text-anchored post-training framework that debiases the shared language decoder without requiring any visual training data. TACT routes evidence-following and prior-driven reasoning trajectories into different optimization stages, enabling the decoder to resolve prior--evidence conflicts. Across counter-commonsense visual benchmarks, TACT substantially improves visual reasoning while preserving general capabilities, demonstrating effective text-to-vision cross-modal transfer.

Related papers

• Staying VIGILant: Mitigating Visual Laziness via Counterfactual Visual Alignment in MLLMs [34.999900039604846]
  Multimodal large language models (MLLMs) extend large language models (LLMs) with visual perception, enabling joint reasoning over images and text.<n>Mechanistic studies indicate that this weakness stems from visual laziness.<n>We propose Visual Information Gain In aLignment (VIGIL), a reinforcement-learning framework that shifts the focus from numerical reward fitting to causal visual grounding.
  arXiv  Detail & Related papers  (2026-06-24T21:12:24Z)
• TVI-CoT: Text-Visual Interleaved Chain-of-Thought Reasoning for Multimodal Understanding [10.402346011516423]
  Chain-of-thought (CoT) reasoning has proven effective for enhancing problem-solving in large language models.<n>Existing CoT approaches suffer from a fundamental limitation: they perform reasoning entirely in text.<n>We propose Text-Visual Interleaved Chain-of-Thought (TVI-CoT) framework that enables explicit interleaving of textual reasoning and visual feature access.
  arXiv  Detail & Related papers  (2026-06-07T05:58:39Z)
• Look on Demand: A Cognitive Scheduling Framework for Visual Evidence Acquisition in Multimodal Reasoning [82.86343313807158]
  We argue that a central challenge is how and when visual evidence is introduced into the reasoning process.<n>Motivated by this insight, we propose CSMR, a multimodal reasoning framework in which a language model controls the reasoning process.
  arXiv  Detail & Related papers  (2026-05-27T08:43:13Z)
• Do Vision-Language Models Truly Perform Vision Reasoning? A Rigorous Study of the Modality Gap [14.392019191368746]
  Reasoning in vision-language models (VLMs) has recently attracted significant attention due to its broad applicability.<n>We introduce CrossMath, a novel multimodal reasoning benchmark designed for controlled cross-modal comparisons.<n>We construct each problem in text-only, image-only, and image+text formats guaranteeing identical task-relevant information.
  arXiv  Detail & Related papers  (2026-04-17T17:15:18Z)
• Render-of-Thought: Rendering Textual Chain-of-Thought as Images for Visual Latent Reasoning [23.364264811510598]
  Chain-of-Thought (CoT) prompting has achieved remarkable success in unlocking the reasoning capabilities of Large Language Models (LLMs)<n>We introduce Render-of-Thought (RoT), the first framework to reify the reasoning chain by rendering textual steps into images.<n>Our method achieves 3-4x token compression and substantial inference acceleration compared to explicit CoT.
  arXiv  Detail & Related papers  (2026-01-21T08:09:25Z)
• DIVER: Dynamic Iterative Visual Evidence Reasoning for Multimodal Fake News Detection [6.225860651499494]
  Multimodal fake news detection is crucial for mitigating adversarial misinformation.<n>We propose DIVER (Dynamic Iterative Visual Evidence Reasoning), a framework grounded in a progressive, evidence-driven reasoning paradigm.<n>Experiments on Weibo, Weibo21, and GossipCop demonstrate that DIVER outperforms state-of-the-art baselines by an average of 2.72%.
  arXiv  Detail & Related papers  (2026-01-12T04:01:33Z)
• Seeing to Act, Prompting to Specify: A Bayesian Factorization of Vision Language Action Policy [59.44168425139687]
  BayesVLA is a Bayesian factorization that decomposes the policy into a visual-action prior, supporting seeing-to-act, and a language-conditioned likelihood, enabling prompt-to-specify.<n>Experiments show superior generalization to unseen instructions, objects, and environments compared to existing methods.
  arXiv  Detail & Related papers  (2025-12-12T01:59:23Z)
• When Thinking Drifts: Evidential Grounding for Robust Video Reasoning [68.75730050161219]
  Chain-of-Thought (CoT) mechanism has enhanced reasoning in text-based tasks.<n>CoT often degrades performance in video reasoning, generating verbose but misleading internal monologues.<n>Visual Evidence Reward (VER) is a reinforcement learning framework that explicitly rewards the generation of reasoning traces that are verifiably grounded in visual evidence.
  arXiv  Detail & Related papers  (2025-10-07T16:03:33Z)
• Seeing Before Reasoning: A Unified Framework for Generalizable and Explainable Fake Image Detection [58.82268659497348]
  We argue that the root of this failure lies in a fundamental mismatch: MLLMs are asked to reason about fakes before they can truly see them.<n>We propose Forensic-Chat, a generalizable, explainable, and still-conversational assistant for fake image detection.
  arXiv  Detail & Related papers  (2025-09-29T20:59:19Z)
• Reasoning-Aligned Perception Decoupling for Scalable Multi-modal Reasoning [95.44766931218896]
  Multi-modal large language models (MLLMs) still lag behind text-based reasoning.<n>We introduce Perception-Reasoning Decoupling, which modularizes the MLLM's reasoning component and makes it easily replaceable.<n>We propose a novel reinforcement learning algorithm called Visual Perception Optimization (VPO) to align the MLLM's perceptual output with the final reasoning task.
  arXiv  Detail & Related papers  (2025-06-05T02:28:07Z)
• A Closer Look at Bias and Chain-of-Thought Faithfulness of Large (Vision) Language Models [58.32070787537946]
  Chain-of-thought (CoT) reasoning enhances performance of large language models.<n>We present the first comprehensive study of CoT faithfulness in large vision-language models.
  arXiv  Detail & Related papers  (2025-05-29T18:55:05Z)
• Retrieval-Based Interleaved Visual Chain-of-Thought in Real-World Driving Scenarios [69.00444996464662]
  We propose RIV-CoT, a Retrieval-Based Interleaved Visual Chain-of-Thought method that enables vision-language models to reason using visual crops corresponding to relevant entities.<n>Our experiments demonstrate that RIV-CoT improves answer accuracy by 3.1% and reasoning accuracy by 4.6% over vanilla CoT prompting.
  arXiv  Detail & Related papers  (2025-01-08T18:31:16Z)
• Fine-Grained Verifiers: Preference Modeling as Next-token Prediction in Vision-Language Alignment [57.0121616203175]
  We propose FiSAO, a novel self-alignment method that utilizes the model's own visual encoder as a fine-grained verifier to improve vision-language alignment.<n>By leveraging token-level feedback from the vision encoder, FiSAO significantly improves vision-language alignment, even surpassing traditional preference tuning methods that require additional data.
  arXiv  Detail & Related papers  (2024-10-18T03:34:32Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.