TARS: MinMax Token-Adaptive Preference Strategy for MLLM Hallucination Reduction

• URL: http://arxiv.org/abs/2507.21584v3
• Date: Sat, 09 Aug 2025 12:17:06 GMT
• Title: TARS: MinMax Token-Adaptive Preference Strategy for MLLM Hallucination Reduction
• Abstract summary: Multimodal large language models (MLLMs) enable vision- reasoning, yet often generate plausible outputs that are factually incorrect or visually ungrounded.<n>Direct preference optimization (DPO) is a common strategy for correcting hallucinations by aligning model outputs with human preferences.<n>We propose TARS, a token-adaptive preference strategy that reformulates DPO as a min-max optimization problem.
• Score: 17.776737597391843
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Multimodal large language models (MLLMs) enable vision-language reasoning, yet often generate plausible outputs that are factually incorrect or visually ungrounded, thereby compromising their reliability. Direct preference optimization (DPO) is a common strategy for correcting hallucinations by aligning model outputs with human preferences. Existing DPO strategies typically treat hallucination-related preferences as fixed targets, relying on static supervision signals during training. This approach tends to overfit to superficial linguistic cues in preference data, leading to distributional rigidity and spurious correlations that impair grounding in causally relevant visual information. To overcome this limitation, we propose TARS, a token-adaptive preference strategy that reformulates DPO as a min-max optimization problem. TARS maximizes token-level distributional shifts under semantic constraints to simulate alignment uncertainty, and simultaneously minimizes the expected preference loss under these controlled perturbations. This joint objective preserves causal grounding while mitigating overfitting to preference patterns, thereby reducing hallucinations in multimodal reasoning. We evaluate TARS on multiple hallucination benchmarks and find consistently strong performance. Using only 4.8k preference samples and no expert feedback, TARS reduces hallucination rates from 26.4% to 13.2% and decreases cognition value from 2.5 to 0.4. It outperforms standard DPO and matches GPT-4o on several key metrics.

Related papers

• Context Blindness in DPO: Mitigating Object Hallucination in MLLMs via Context-Calibrated Preference Optimization [34.25772615253414]
  We propose a metric that measures how much a model's preference strengthens when relevant context is provided.<n>We find that higher CPG consistently corresponds to lower hallucination, yet standard DPO and its variants exhibit only limited CPG.<n>We propose Context-Calibrated DPO (C$2$-DPO), which directly maximizes CPG while preserving the original preference ordering.
  arXiv  Detail & Related papers  (2026-08-12T15:20:29Z)
• S-SPPO: Semantic-Calibrated Self-Play Preference Optimization [36.01916066772865]
  We introduce Self-Play Preference Optimization (SPPO), which iteratively refines the policy by training on self-generated win-lose pairs.<n>Our investigation reveals a critical instability in SPPO: the optimization is prone to policy degeneration when the preference assigns overly confident wins to semantically indistinguishable responses.<n>We show that S-SPPO avoids the performance degradation seen in prior methods, achieving 52.19% win rate and 47.46% length-controlled win rate on AlpacaEval 2.0 with Llama-3-8B, without using additional human-annotated preferences during training.
  arXiv  Detail & Related papers  (2026-06-01T02:06:58Z)
• wDPO: Winsorized Direct Preference Optimization for Robust LLM Alignment [48.487557157323664]
  In practice, preference data are often noisy.<n>Existing robust variants of DPO mainly rely on uniform objective modifications or global reweighting.<n>We show that robust preference alignment benefits from addressing different noise types with targeted interventions.
  arXiv  Detail & Related papers  (2026-03-07T13:30:53Z)
• MoD-DPO: Towards Mitigating Cross-modal Hallucinations in Omni LLMs using Modality Decoupled Preference Optimization [4.088161686930475]
  We propose Modality-Decoupled Direct Preference Optimization (MoD-DPO) for improving modality grounding in omni LLMs.<n>MoD-DPO introduces modality-aware regularization terms that explicitly enforce invariance to corruptions in irrelevant modalities and sensitivity to perturbations in relevant modalities.<n>Experiments demonstrate that MoD-DPO consistently improves perception accuracy and hallucination resistance, outperforming previous preference optimization baselines.
  arXiv  Detail & Related papers  (2026-03-03T17:50:24Z)
• DA-DPO: Cost-efficient Difficulty-aware Preference Optimization for Reducing MLLM Hallucinations [22.299736215070343]
  Multimodal Large Language Models (MLLMs) tend to overemphasize easily distinguishable preference pairs.<n>We propose Difficulty-Aware Direct Preference Optimization (DA-DPO), a cost-effective framework designed to balance the learning process.
  arXiv  Detail & Related papers  (2026-01-02T09:41:54Z)
• Token-Level Inference-Time Alignment for Vision-Language Models [58.41370989069588]
  Vision-Language Models (VLMs) have become essential backbones of modern multimodal intelligence.<n>We present TITA, a lightweight framework that freezes the base VLM and instead trains a reward model to approximate its distribution.<n>During inference, implicit preference signals are extracted as log-probability ratios between the reward model and the target VLM, yielding dense autoregressive feedback.
  arXiv  Detail & Related papers  (2025-10-20T09:58:03Z)
• From Noisy Traces to Stable Gradients: Bias-Variance Optimized Preference Optimization for Aligning Large Reasoning Models [90.45197506653341]
  Large reasoning models generate intermediate reasoning traces before producing final answers.<n> aligning LRMs with human preferences, a crucial prerequisite for model deployment, remains underexplored.<n>A common workaround optimized a single sampled trajectory, which introduces substantial gradient variance from trace sampling.
  arXiv  Detail & Related papers  (2025-10-06T17:58:01Z)
• Latent Collective Preference Optimization: A General Framework for Robust LLM Alignment [7.1259212876994695]
  We introduce Latent Collective Preference Optimization (LCPO) to learn the latent collective consensus from noisy data.<n>Our experiments demonstrate LCPO's effectiveness as a general framework, consistently enhancing four state-of-the-art alignment algorithms.<n>When applied to Mistral and Llama 3 models, LCPO-enhanced methods achieve substantial win rate gains on AlpacaEval 2 and Arena-Hard, with improvements of up to 7.0% on both benchmarks.
  arXiv  Detail & Related papers  (2025-09-29T01:17:49Z)
• Mitigating Hallucination Through Theory-Consistent Symmetric Multimodal Preference Optimization [58.64721525687295]
  Direct Preference Optimization (DPO) has emerged as an effective approach for mitigating hallucination in Multimodal Large Language Models (MLLMs)<n>We propose a Symmetric Multimodal Preference Optimization (SymMPO) which conducts symmetric preference learning with direct preference supervision (i.e., response pairs)<n>In addition to conventional ordinal preference learning, SymMPO introduces a preference margin consistency loss to quantitatively regulate the preference gap between symmetric preference pairs.
  arXiv  Detail & Related papers  (2025-06-13T12:29:15Z)
• CHiP: Cross-modal Hierarchical Direct Preference Optimization for Multimodal LLMs [107.21334626890713]
  Multimodal Large Language Models (MLLMs) still struggle with hallucinations despite their impressive capabilities.<n>We propose a Cross-modal Hierarchical Direct Preference Optimization (CHiP) to address these limitations.<n>We evaluate CHiP through both quantitative and qualitative analyses, with results across multiple benchmarks demonstrating its effectiveness in reducing hallucinations.
  arXiv  Detail & Related papers  (2025-01-28T02:05:38Z)
• Mitigating Hallucinations in Large Vision-Language Models via DPO: On-Policy Data Hold the Key [24.229983103296988]
  Hallucination remains a major challenge for Large Vision-Language Models (LVLMs)<n>We propose On-Policy Alignment (OPA)-DPO framework, which uniquely leverages expert feedback to correct hallucinated responses.<n>OPA-DPO achieves an additional reduction in the hallucination rate of LLaVA-1.5-7B: 13.26% on the AMBER benchmark and 5.39% on the Object-Hal benchmark.
  arXiv  Detail & Related papers  (2025-01-16T17:48:03Z)
• Uncertainty-Penalized Direct Preference Optimization [52.387088396044206]
  We develop a pessimistic framework for DPO by introducing preference uncertainty penalization schemes. The penalization serves as a correction to the loss which attenuates the loss gradient for uncertain samples. We show improved overall performance compared to vanilla DPO, as well as better completions on prompts from high-uncertainty chosen/rejected responses.
  arXiv  Detail & Related papers  (2024-10-26T14:24:37Z)
• Revisiting Essential and Nonessential Settings of Evidential Deep Learning [70.82728812001807]
  Evidential Deep Learning (EDL) is an emerging method for uncertainty estimation. We propose Re-EDL, a simplified yet more effective variant of EDL.
  arXiv  Detail & Related papers  (2024-10-01T04:27:07Z)
• Geometric-Averaged Preference Optimization for Soft Preference Labels [78.2746007085333]
  Many algorithms for aligning LLMs with human preferences assume that human preferences are binary and deterministic.<n>In this work, we introduce the distributional soft preference labels and improve Direct Preference Optimization (DPO) with a weighted geometric average of the LLM output likelihood in the loss function.
  arXiv  Detail & Related papers  (2024-09-10T17:54:28Z)
• Self-Play Preference Optimization for Language Model Alignment [75.83359213697854]
  Recent advancements suggest that directly working with preference probabilities can yield a more accurate reflection of human preferences. We propose a self-play-based method for language model alignment, which treats the problem as a constant-sum two-player game. Our approach, dubbed Self-Play Preference Optimization (SPPO), utilizes iterative policy updates to provably approximate the Nash equilibrium.
  arXiv  Detail & Related papers  (2024-05-01T17:59:20Z)
• Dissecting Human and LLM Preferences [80.55271307662365]
  We find that humans are less sensitive to errors, favor responses that support their stances, and show clear dislike when models admit their limits. advanced LLMs like GPT-4-Turbo emphasize correctness, clarity, and harmlessness more. We show that preference-based evaluation can be intentionally manipulated.
  arXiv  Detail & Related papers  (2024-02-17T14:34:31Z)
• Beyond Hallucinations: Enhancing LVLMs through Hallucination-Aware Direct Preference Optimization [45.53216822981202]
  This paper introduces a novel solution, Hallucination-Aware Direct Preference Optimization (HA-DPO), which reframes the hallucination problem as a preference selection task. When applied to three mainstream multimodal models, HA-DPO significantly reduced hallucination issues and amplified the models' generalization capabilities.
  arXiv  Detail & Related papers  (2023-11-28T14:54:37Z)
• Statistical Rejection Sampling Improves Preference Optimization [42.57245965632205]
  We introduce a novel approach to source preference data from the target optimal policy using rejection sampling. We also propose a unified framework that enhances the loss functions used in both Sequence Likelihood (SLiC) and Direct Preference Optimization (DPO) from a preference modeling standpoint.
  arXiv  Detail & Related papers  (2023-09-13T01:07:25Z)

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.