P2R decouples perception from reasoning in VLMs via a two-stage process and PRA-GRPO alternating RL training, reporting gains such as 93.2% on V-Star for the 4B model over its Qwen3-VL backbone.
hub Canonical reference
Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search
Canonical reference. 71% of citing Pith papers cite this work as background.
35
Pith papers citing it
Background
71% of classified citations
abstract
Recent advances in large multimodal models have leveraged image-based tools with reinforcement learning to tackle visual problems. However, existing open-source approaches often exhibit monotonous reasoning patterns and allow only a limited number of interaction turns, making them inadequate for difficult tasks that require trial-and-error exploration. In this work, we address this limitation by scaling up tool-based interactions and introduce Mini-o3, a system that executes deep, multi-turn reasoning -- spanning tens of steps -- and achieves state-of-the-art performance on challenging visual search tasks. Our recipe for reproducing OpenAI o3-style behaviors comprises three key components. First, we construct the Visual Probe Dataset, a collection of thousands of challenging visual search problems designed for exploratory reasoning. Second, we develop an iterative data collection pipeline to obtain cold-start trajectories that exhibit diverse reasoning patterns, including depth-first search, trial-and-error, and goal maintenance. Third, we propose an over-turn masking strategy that prevents penalization of over-turn responses (those that hit the maximum number of turns) during reinforcement learning, thereby balancing training-time efficiency with test-time scalability. Despite training with an upper bound of only six interaction turns, our model generates trajectories that naturally scale to tens of turns at inference time, with accuracy improving as the number of turns increases. Extensive experiments demonstrate that Mini-o3 produces rich reasoning patterns and deep thinking paths, effectively solving challenging visual search problems.
hub tools
citation-role summary
background 5
baseline 1
dataset 1
citation-polarity summary
representative citing papers
DiCoBench is a new high-resolution multi-image benchmark exposing large gaps between top MLLMs and human performance (98.3%) on differential and commonality visual cue perception.
Astra couples an RL-trained VLM policy with a view-consistent Bagel-based world simulator to enable agentic imagination during spatial reasoning, yielding benchmark gains on MMSI-Bench and MindCube.
DeepLatent introduces a parallel latent visual reasoning framework with learnable 2D tokens and continuous RL, trained via distillation then RL, plus a new 180K dataset, claiming SOTA benchmark results.
CaST-Bench creates a benchmark with causal-chain annotations and novel metrics showing that current VLMs struggle to construct precise grounded causal chains in video QA.
Speaker-Reasoner is an end-to-end speech LLM that iteratively analyzes audio structure, predicts temporal boundaries, and jointly models speaker identity, gender, timestamps, and transcription using a speaker-aware cache for long audio.
VideoThinker uses LLM-generated synthetic tool trajectories in caption space grounded to video frames to train agentic VideoLLMs that outperform baselines on long-video benchmarks.
PixelEyes decouples reasoning and perception via mask-guided search and semantic BFS, introduces PixelEyes-6K dataset and Pinpoint-Bench benchmark, and open-sources code and models.
VisReflect generates continuous latent visual reflections to emphasize relevant visual features and guide attention in LVLMs, yielding 4.1% gains on image benchmarks and 1.8% on video benchmarks with 44% less inference time than zooming methods.
AXPO addresses the Thinking-Acting Gap in agentic RL training by targeted resampling of tool calls in all-wrong subgroups, delivering +1.8pp gains over GRPO on nine multimodal benchmarks with an 8B model beating a 32B baseline on Pass@4.
SeProD is a plug-and-play self-prophetic decoding framework that combines pre- and post-training LVLM capabilities via probability-based sampling to improve coherent visual search and multi-step reasoning.
InterSketch improves long-horizon visual-textual chain-of-thought in VLMs by dynamically generating and interleaving self-correcting visual sketches with text, using a synthesized dataset plus reflection in cold-start followed by stepwise-reward RL, and reports outperforming Gemini-3-Pro on benchmar
AutoTool uses dual-mode RL to let MLLMs adaptively choose tool use or text-only reasoning, reporting 21.8% accuracy gain on V* and 44.9% efficiency gain on POPE versus baselines.
Vision-OPD transfers an MLLM's privileged regional perception to its full-image policy through on-policy token-level self-distillation, yielding competitive results on fine-grained visual benchmarks.
Constraining visual token budgets during SFT and RL forces VLMs to learn functional active perception, yielding ~5% relative gains and strong transfer to unconstrained evaluation.
Imagining in 360° decouples visual search into a single-step probabilistic semantic layout predictor and an actor, removing the need for multi-turn CoT reasoning and trajectory annotations while improving efficiency in 360° environments.
HierVA improves multi-step chart question answering by having a high-level manager maintain key joint contexts while specialized workers perform targeted reasoning with visual zoom-in.
ToolsRL trains MLLMs via a tool-specific then accuracy-focused RL curriculum to master visual tools for complex reasoning tasks.
Q-Zoom achieves up to 4.39x inference speedup in high-resolution MLLM scenarios via query-aware gating and region localization, matching or exceeding baseline accuracy on document and high-res benchmarks.
MAPO improves multimodal chain-of-thought reasoning by requiring explicit textual descriptions of visual tool results and using a novel advantage estimator that combines semantic alignment with task rewards.
CharTool equips MLLMs with cropping and code tools plus agentic RL on DuoChart data to raise chart-reasoning accuracy by up to 9.78 percent on benchmarks.
AdaTooler-V trains MLLMs to adaptively use vision tools via AT-GRPO reinforcement learning and new datasets, reaching 89.8% on V* and outperforming GPT-4o.
Skyra is an MLLM that detects AI-generated videos by identifying and reasoning over grounded visual artifacts, supported by a new annotated dataset and benchmark.
Activation Replay boosts multimodal reasoning in post-trained LMMs by replaying low-entropy activations from base models to RLVR counterparts at test time via visual token manipulation.
citing papers explorer
-
Perceive-to-Reason: Decoupling Perception and Reasoning for Fine-Grained Visual Reasoning
P2R decouples perception from reasoning in VLMs via a two-stage process and PRA-GRPO alternating RL training, reporting gains such as 93.2% on V-Star for the 4B model over its Qwen3-VL backbone.
-
DiCoBench: Benchmarking Multi-Image Fine-Grained Perception via Differential and Commonality Visual Cues
DiCoBench is a new high-resolution multi-image benchmark exposing large gaps between top MLLMs and human performance (98.3%) on differential and commonality visual cue perception.
-
Thinking with Imagination: Agentic Visual Spatial Reasoning with World Simulators
Astra couples an RL-trained VLM policy with a view-consistent Bagel-based world simulator to enable agentic imagination during spatial reasoning, yielding benchmark gains on MMSI-Bench and MindCube.
-
DeepLatent: Think with Images via Parallel Latent Visual Reasoning
DeepLatent introduces a parallel latent visual reasoning framework with learnable 2D tokens and continuous RL, trained via distillation then RL, plus a new 180K dataset, claiming SOTA benchmark results.
-
CaST-Bench: Benchmarking Causal Chain-Grounded Spatio-Temporal Reasoning for Video Question Answering
CaST-Bench creates a benchmark with causal-chain annotations and novel metrics showing that current VLMs struggle to construct precise grounded causal chains in video QA.
-
Speaker-Reasoner: Scaling Interaction Turns and Reasoning Patterns for Timestamped Speaker-Attributed ASR
Speaker-Reasoner is an end-to-end speech LLM that iteratively analyzes audio structure, predicts temporal boundaries, and jointly models speaker identity, gender, timestamps, and transcription using a speaker-aware cache for long audio.
-
VideoThinker: Building Agentic VideoLLMs with LLM-Guided Tool Reasoning
VideoThinker uses LLM-generated synthetic tool trajectories in caption space grounded to video frames to train agentic VideoLLMs that outperform baselines on long-video benchmarks.
-
PixelEyes: Decoupling Perception and Reasoning for Pinpoint Visual Evidence Seeking
PixelEyes decouples reasoning and perception via mask-guided search and semantic BFS, introduces PixelEyes-6K dataset and Pinpoint-Bench benchmark, and open-sources code and models.
-
VisReflect: Latent Visual Reflection for Fine-Grained Perception in Long Visual Context
VisReflect generates continuous latent visual reflections to emphasize relevant visual features and guide attention in LVLMs, yielding 4.1% gains on image benchmarks and 1.8% on video benchmarks with 44% less inference time than zooming methods.
-
Agent Explorative Policy Optimization for Multimodal Agentic Reasoning
AXPO addresses the Thinking-Acting Gap in agentic RL training by targeted resampling of tool calls in all-wrong subgroups, delivering +1.8pp gains over GRPO on nine multimodal benchmarks with an 8B model beating a 32B baseline on Pass@4.
-
Self-Prophetic Decoding to Unlock Visual Search in LVLMs
SeProD is a plug-and-play self-prophetic decoding framework that combines pre- and post-training LVLM capabilities via probability-based sampling to improve coherent visual search and multi-step reasoning.
-
InterSketch: An Interleaved Reasoning Model with Self-correcting Visual Sketch and Stepwise Reward
InterSketch improves long-horizon visual-textual chain-of-thought in VLMs by dynamically generating and interleaving self-correcting visual sketches with text, using a synthesized dataset plus reflection in cold-start followed by stepwise-reward RL, and reports outperforming Gemini-3-Pro on benchmar
-
Are Tools Always Beneficial? Learning to Invoke Tools Adaptively for Dual-Mode Multimodal LLM Reasoning
AutoTool uses dual-mode RL to let MLLMs adaptively choose tool use or text-only reasoning, reporting 21.8% accuracy gain on V* and 44.9% efficiency gain on POPE versus baselines.
-
Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation
Vision-OPD transfers an MLLM's privileged regional perception to its full-image policy through on-policy token-level self-distillation, yielding competitive results on fine-grained visual benchmarks.
-
Starve to Perceive: Taming Lazy Perception in VLMs with Constrained Visual Bandwidth
Constraining visual token budgets during SFT and RL forces VLMs to learn functional active perception, yielding ~5% relative gains and strong transfer to unconstrained evaluation.
-
Beyond Thinking: Imagining in 360$^\circ$ for Humanoid Visual Search
Imagining in 360° decouples visual search into a single-step probabilistic semantic layout predictor and an actor, removing the need for multi-turn CoT reasoning and trajectory annotations while improving efficiency in 360° environments.
-
Hierarchical Visual Agent: Managing Contexts in Joint Image-Text Space for Advanced Chart Reasoning
HierVA improves multi-step chart question answering by having a high-level manager maintain key joint contexts while specialized workers perform targeted reasoning with visual zoom-in.
-
Visual Reasoning through Tool-supervised Reinforcement Learning
ToolsRL trains MLLMs via a tool-specific then accuracy-focused RL curriculum to master visual tools for complex reasoning tasks.
-
Q-Zoom: Query-Aware Adaptive Perception for Efficient Multimodal Large Language Models
Q-Zoom achieves up to 4.39x inference speedup in high-resolution MLLM scenarios via query-aware gating and region localization, matching or exceeding baseline accuracy on document and high-res benchmarks.
-
Walk the Talk: Bridging the Reasoning-Action Gap for Thinking with Images via Multimodal Agentic Policy Optimization
MAPO improves multimodal chain-of-thought reasoning by requiring explicit textual descriptions of visual tool results and using a novel advantage estimator that combines semantic alignment with task rewards.
-
CharTool: Tool-Integrated Visual Reasoning for Chart Understanding
CharTool equips MLLMs with cropping and code tools plus agentic RL on DuoChart data to raise chart-reasoning accuracy by up to 9.78 percent on benchmarks.
-
AdaTooler-V: Adaptive Tool-Use for Images and Videos
AdaTooler-V trains MLLMs to adaptively use vision tools via AT-GRPO reinforcement learning and new datasets, reaching 89.8% on V* and outperforming GPT-4o.
-
Skyra: AI-Generated Video Detection via Grounded Artifact Reasoning
Skyra is an MLLM that detects AI-generated videos by identifying and reasoning over grounded visual artifacts, supported by a new annotated dataset and benchmark.
-
Boosting Reasoning in Large Multimodal Models via Activation Replay
Activation Replay boosts multimodal reasoning in post-trained LMMs by replaying low-entropy activations from base models to RLVR counterparts at test time via visual token manipulation.
-
CropVLM: Learning to Zoom for Fine-Grained Vision-Language Perception
CropVLM uses reinforcement learning to learn image zooming policies that boost fine-grained perception in VLMs on out-of-domain high-resolution tasks without labeled boxes, synthetic data, or VLM changes.
-
DeepEyesV2: Toward Agentic Multimodal Model
DeepEyesV2 uses a two-stage cold-start plus reinforcement learning pipeline to produce an agentic multimodal model that adaptively invokes tools and outperforms direct RL on real-world reasoning benchmarks.
-
V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning
V-Zero trains MLLMs for visual reasoning without answer labels by gating on-policy distillation trajectories using contrastive evidence from relevant versus negative image crops.
-
Latent Visual States for Efficient Multimodal Reasoning
EVA generates adaptive Latent_slot tokens as internal visual thoughts, trained end-to-end with text tokens via D-GSPO on the EVA-230K dataset, claiming performance gains and better inference efficiency.
-
CaptchaMind: Training CAPTCHA Solvers via Reinforcement Learning with Explicit Reasoning Supervision
Presents CaptchaBench benchmark and CaptchaMind RL solver achieving 82.9% success on benchmark tasks and 71% on real-world CAPTCHAs via explicit reasoning process supervision.
-
Q-DeepSight: Incentivizing Thinking with Images for Image Quality Assessment and Refinement
Q-DeepSight proposes a think-with-image multimodal CoT framework trained via RL with perceptual curriculum rewards and evidence gradient filtering to achieve SOTA IQA performance and enable training-free perceptual refinement in image generation.
-
HiVLA: A Visual-Grounded-Centric Hierarchical Embodied Manipulation System
HiVLA decouples VLM-based semantic planning with visual grounding from a cascaded cross-attention DiT action expert, outperforming end-to-end VLAs on long-horizon and fine-grained manipulation.
-
Act Wisely: Cultivating Meta-Cognitive Tool Use in Agentic Multimodal Models
HDPO reframes tool efficiency as a conditional objective within accurate trajectories, enabling Metis to reduce tool invocations by orders of magnitude while raising reasoning accuracy.
-
Learning to Focus and Precise Cropping: A Reinforcement Learning Framework with Information Gaps and Grounding Loss for MLLMs
A two-stage RL method with information gaps and grounding loss trains MLLMs to focus on and precisely crop relevant image regions, yielding SOTA results on high-resolution VQA benchmarks.
-
UniReason-Med: A Shared Grounded Reasoning Interface for 2D-to-3D Transfer in Medical VQA
UniReason-Med introduces a unified framework for 2D and 3D medical VQA with shared grounded reasoning, trained on a 220K dataset, claiming that joint 2D+3D supervision improves 3D performance over 3D-only training.
-
Kwai Keye-VL-2.0 Technical Report
Kwai Keye-VL-2.0-30B-A3B is a 30B MoE model with 3B active parameters using DSA adaptation and MOPD distillation that reports SOTA results on video understanding and agent benchmarks.