arXiv:2609.21675v1 Announce Type: new Abstract: Despite the remarkable progress in Multimodal Large Language Models (MLLMs), prevailing Chain-of-Thought (CoT) paradigms remain confined to the natural...
The paper introduces V‑Rubrics, a reinforcement‑learning framework that evaluates vision‑language model responses by breaking them into atomic propositions and scoring them on Visual Faithfulness, Reasoning Consistency, and Instruction Following. Using a fine‑tuned Qwen3‑VL‑8B‑Instruct model and a newly created 50K‑example V‑Rubrics dataset, the authors demonstrate that rubric‑based GRPO outperforms both a shared SFT baseline and an answer‑only GRPO, especially on knowledge‑oriented and visually grounded reasoning tasks.
The paper introduces PIVOT, a dual-level learning framework designed to improve visually-grounded multimodal reasoning in large vision-language models. PIVOT employs a self‑calibrated experience replay mechanism to selectively reuse valuable visual reasoning trajectories, and a vision‑guided advantage allocation scheme that assigns extra rewards to tokens with strong visual support. Experiments on multiple benchmarks show that PIVOT enhances the multimodal reasoning performance of these models.
arXiv:2609.39168v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has improved the reasoning capabilities of Multimodal Large Language Models (MLLMs), yet existing...
GeoReform is a reflective formalization evolution framework designed to improve multimodal geometry problem solving by treating formalization as an optimizable policy rather than a fixed parser. It iteratively runs the reasoning pipeline, collects failures, diagnoses representation defects, and mutates the policy to better select, ground, group, and present geometric entities, relations, constraints, and targets. On the Geometry3K benchmark, GeoReform raises Qwen3VL-2B accuracy from 42.0% to 56.0%, demonstrating that effective formalization is crucial for enhancing multimodal geometry reasoning.
arXiv:2604. 04917v3 Announce Type: replace-cross Abstract: What does it take to build a visual reasoner that works across charts, science, spatial understanding, and open-ended tasks?
The paper introduces Numerical Visual Chain-of-Thought (NV‑CoT), a framework that expands multimodal large language models (MLLMs) to reason over images using continuous numerical coordinates instead of discrete tokens or fixed patches. NV‑CoT allows MLLMs to generate bounding‑box coordinates directly as actions with minimal architectural changes, supporting both supervised fine‑tuning and reinforcement learning via a Gaussian or Laplace policy. Experiments on three benchmarks show that NV‑CoT improves localization precision, answer accuracy, and training convergence compared to 11 visual reasoning baselines.
arXiv:2608. 07581v1 Announce Type: cross Abstract: Group-based reinforcement learning methods for multimodal large language models typically rely on trajectory-level credit assignment that applies a single advantage to all tokens in a response.
The paper introduces RECAP, a redundancy-aware credit assignment method that improves reasoning efficiency in large language models by assigning credit to each reasoning step based on its downstream role and contribution to the correct answer. RECAP uses a semantic dependency graph to measure structural responsibility and evaluates step efficacy via changes in gold-answer log-likelihood, enabling step-specific updates without requiring a separate reward model or concise trajectories. Experiments on two 7B models across four mathematical reasoning benchmarks show that RECAP enhances the accuracy-efficiency trade-off, boosting pass@1 by 2.0–3.7 percentage points while cutting reasoning tokens by 8–31% compared to GRPO.
SLVR (Structured Latent Visual Reasoning) is a training framework that blends explicit chain-of-thought with latent reasoning for multimodal large language models. It organizes reasoning into typed latent stages—planning, grounding, evidence selection, and integration—each supervised by corresponding signals such as plans, bounding boxes, visual evidence, and final rationales. Built on Qwen2.5‑VL‑7B, SLVR consistently improves performance on multimodal reasoning benchmarks, achieving significant gains on MMVP, BLINK Relation, V*, MathVista, and ChartQA.
arXiv:2601.08010v4 Announce Type: replace Abstract: Vision-language models achieve strong performance across a wide range of multimodal understanding and reasoning tasks, yet their multi-step reasoni...
arXiv:2607. 03748v1 Announce Type: new Abstract: Unified multi-modal models (UMMs) have shown promising interleaved text-image reasoning capabilities, yet effectively optimizing such multi-turn generation via reinforcement learning (RL) remains an open challenge.