Controlling Multimodal LLMs via Reward-guided Decoding

• URL: http://arxiv.org/abs/2508.11616v1
• Date: Fri, 15 Aug 2025 17:29:06 GMT
• Title: Controlling Multimodal LLMs via Reward-guided Decoding
• Authors: Oscar Mañas, Pierluca D'Oro, Koustuv Sinha, Adriana Romero-Soriano, Michal Drozdzal, Aishwarya Agrawal,
• Abstract summary: We study the adaptation of Multimodal Large Language Models (MLLMs) through controlled decoding.<n>Our method involves building reward models for visual grounding and using them to guide the MLLM's decoding process.<n>We evaluate our method on standard object hallucination benchmarks, showing that it provides significant controllability over MLLM inference.
• Score: 17.5544679985101
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: As Multimodal Large Language Models (MLLMs) gain widespread applicability, it is becoming increasingly desirable to adapt them for diverse user needs. In this paper, we study the adaptation of MLLMs through controlled decoding. To achieve this, we introduce the first method for reward-guided decoding of MLLMs and demonstrate its application in improving their visual grounding. Our method involves building reward models for visual grounding and using them to guide the MLLM's decoding process. Concretely, we build two separate reward models to independently control the degree of object precision and recall in the model's output. Our approach enables on-the-fly controllability of an MLLM's inference process in two ways: first, by giving control over the relative importance of each reward function during decoding, allowing a user to dynamically trade off object precision for recall in image captioning tasks; second, by giving control over the breadth of the search during decoding, allowing the user to control the trade-off between the amount of test-time compute and the degree of visual grounding. We evaluate our method on standard object hallucination benchmarks, showing that it provides significant controllability over MLLM inference, while consistently outperforming existing hallucination mitigation methods.

Related papers

• A Comprehensive Study on Visual Token Redundancy for Discrete Diffusion-based Multimodal Large Language Models [85.30893355216486]
  We study how visual token redundancy evolves with different dMLLM architectures and tasks.<n>Our study reveals that visual redundancy emerges only in from-scratch dMLLMs while handling long-answer tasks.<n>Layer-skipping is promising for accelerating AR-to-diffusion dMLLMs, whereas progressive or late-step pruning is more effective for from-scratch dMLLMs.
  arXiv  Detail & Related papers  (2025-11-19T04:13:36Z)
• Multimodal LLMs as Customized Reward Models for Text-to-Image Generation [60.164968941945645]
  We introduce LLaVA-Reward, an efficient reward model designed to automatically evaluate text-to-image (T2I) generations across multiple perspectives.<n>LLaVA-Reward directly utilizes the hidden states of multimodal large language models (MLLMs)<n>We train LLaVA-Reward on four evaluation perspectives: text-image alignment, fidelity/artifact, safety, and overall ranking.
  arXiv  Detail & Related papers  (2025-07-28T23:52:53Z)
• Vad-R1: Towards Video Anomaly Reasoning via Perception-to-Cognition Chain-of-Thought [58.321044666612174]
  Vad-R1 is an end-to-end MLLM-based framework for Video Anomaly Reasoning.<n>We design a Perception-to-Cognition Chain-of-Thought (P2C-CoT) that simulates the human process of recognizing anomalies.<n>We also propose an improved reinforcement learning algorithm AVA-GRPO, which explicitly incentivizes the anomaly reasoning capability of MLLMs.
  arXiv  Detail & Related papers  (2025-05-26T12:05:16Z)
• Instance-Level Data-Use Auditing of Visual ML Models [49.862257986549885]
  Growing trend of legal disputes over the unauthorized use of data in machine learning (ML) systems highlights the need for reliable data-use auditing mechanisms.<n>We present the first proactive, instance-level, data-use auditing method designed to enable data owners to audit the use of their individual data instances in ML models.
  arXiv  Detail & Related papers  (2025-03-28T13:28:57Z)
• LLaVA Steering: Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steering [30.51487692912812]
  Multimodal Large Language Models (MLLMs) have significantly advanced visual tasks by integrating visual representations into large language models (LLMs)<n>We introduce Modality Linear Representation-Steering (MoReS) to achieve the goal.<n>MoReS effectively re-balances the intrinsic modalities throughout the model, where the key idea is to steer visual representations through linear transformations in the visual subspace across each model layer.
  arXiv  Detail & Related papers  (2024-12-16T21:14:11Z)
• OLA-VLM: Elevating Visual Perception in Multimodal LLMs with Auxiliary Embedding Distillation [95.78870389271832]
  The standard practice for developing contemporary MLLMs is to feed features from vision encoder(s) into the LLM and train with natural language supervision.<n>We propose OLA-VLM, the first approach distilling knowledge into the LLM's hidden representations from a set of target visual representations.<n>We show that OLA-VLM boosts performance by an average margin of up to 2.5% on various benchmarks, with a notable improvement of 8.7% on the Depth task in CV-Bench.
  arXiv  Detail & Related papers  (2024-12-12T18:55:18Z)
• LLaVA-KD: A Framework of Distilling Multimodal Large Language Models [72.68665884790002]
  We propose a novel framework to transfer knowledge from l-MLLMs to s-MLLMs.<n>We introduce Multimodal Distillation (MDist) to transfer teacher model's robust representations across both visual and linguistic modalities.<n>We also propose a three-stage training scheme to fully exploit the potential of the proposed distillation strategy.
  arXiv  Detail & Related papers  (2024-10-21T17:41:28Z)
• CoMMIT: Coordinated Instruction Tuning for Multimodal Large Language Models [68.64605538559312]
  In this paper, we analyze the MLLM instruction tuning from both theoretical and empirical perspectives. Inspired by our findings, we propose a measurement to quantitatively evaluate the learning balance. In addition, we introduce an auxiliary loss regularization method to promote updating of the generation distribution of MLLMs.
  arXiv  Detail & Related papers  (2024-07-29T23:18:55Z)
• Aligning Large Language Models with Representation Editing: A Control Perspective [38.71496554018039]
  Fine-tuning large language models (LLMs) to align with human objectives is crucial for real-world applications. Test-time alignment techniques, such as prompting and guided decoding, do not modify the underlying model. We propose aligning LLMs through representation editing.
  arXiv  Detail & Related papers  (2024-06-10T01:21:31Z)

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.