EgoGapBench shows humans reliably select egocentric actions in multi-agent scenes while MLLMs systematically choose other agents' actions, and standard egocentric training data fails to close the gap.
super hub Mixed citations
Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling
Mixed citation behavior. Most common role is background (56%).
261
Pith papers citing it
14
external citations · Pith
Background
56% of classified citations
abstract
We introduce InternVL 2.5, an advanced multimodal large language model (MLLM) series that builds upon InternVL 2.0, maintaining its core model architecture while introducing significant enhancements in training and testing strategies as well as data quality. In this work, we delve into the relationship between model scaling and performance, systematically exploring the performance trends in vision encoders, language models, dataset sizes, and test-time configurations. Through extensive evaluations on a wide range of benchmarks, including multi-discipline reasoning, document understanding, multi-image / video understanding, real-world comprehension, multimodal hallucination detection, visual grounding, multilingual capabilities, and pure language processing, InternVL 2.5 exhibits competitive performance, rivaling leading commercial models such as GPT-4o and Claude-3.5-Sonnet. Notably, our model is the first open-source MLLMs to surpass 70% on the MMMU benchmark, achieving a 3.7-point improvement through Chain-of-Thought (CoT) reasoning and showcasing strong potential for test-time scaling. We hope this model contributes to the open-source community by setting new standards for developing and applying multimodal AI systems. HuggingFace demo see https://huggingface.co/spaces/OpenGVLab/InternVL
hub tools
citation-role summary
background 30
baseline 18
method 4
dataset 1
other 1
citation-polarity summary
claims ledger
- abstract We introduce InternVL 2.5, an advanced multimodal large language model (MLLM) series that builds upon InternVL 2.0, maintaining its core model architecture while introducing significant enhancements in training and testing strategies as well as data quality. In this work, we delve into the relationship between model scaling and performance, systematically exploring the performance trends in vision encoders, language models, dataset sizes, and test-time configurations. Through extensive evaluations on a wide range of benchmarks, including multi-discipline reasoning, document understanding, mult
authors
co-cited works
representative citing papers
MLLMs drop from over 85% accuracy on action presence to under 50% on matched action-denial videos, exposing a causal verification gap that causal graph prompts partially close.
MuseBench shows state-of-the-art MLLMs achieve only 48.29% accuracy on intent-level audiovisual arts understanding versus 87.18% for human experts.
Introduces VG-GUIBench benchmark and TASKER keyframe extraction algorithm that improves performance on VideoQA and video-guided agentic tasks.
Introduces the FeVOS task, a 968-clip dataset with foresight expressions, and an MLLM model FeVOS-R1 trained via SFT then RL that reports SOTA on the new task plus generalization to prior RVOS benchmarks.
Processed egocentric human video outperforms teleoperated real-robot trajectories as pretraining data for embodied foundation models, delivering 24% lower validation loss and 52.5-90% higher task success rates under matched post-training protocols.
Remote sensing MLLMs perform poorly on negation tasks with hallucinations and accuracy drops, but the NeFo test-time learning method substantially improves negation understanding and generalizes to unseen tasks using ~5% unlabeled test samples.
Earth-OneVision is a unified 2B-parameter RS-MLLM supporting six modalities and nine tasks via FGVLA, SLIS, and PCMA mechanisms plus a 34M QA-pair dataset, reporting competitive or superior benchmark results versus larger models.
TVI-CoT introduces learnable control tokens <THINK>, <LOOK>, <ANSWER> that let multimodal LLMs interleave textual reasoning with dynamic visual feature access, reporting gains of 3.4-6.1% on eight benchmarks over prior CoT baselines.
Iterative solvers in layer-wise model merging act as spectral regularizers on an ill-posed interference operator; closed-form SWUDI and adaptive SWUDI-A match or exceed SOTA merging accuracy with 28-72x wall-clock speedup.
SVHighlights is the first benchmark for highlight detection in hour-long sports videos, with TF-SELECTOR showing that segment-level LLM scoring outperforms adapted short-video baselines by 2.5-4 points on key metrics.
DisasterBench is a new multi-stage multimodal reasoning benchmark for UAV disaster response with 14 scenes and 9 tasks; the accompanying 2B DisasterVL model outperforms open-source MLLMs and approaches GPT-4o efficiency.
Authors create ReasonMatch-Bench and DCRL training to boost MLLM performance on wide-baseline matching, reporting gains over baselines while preserving general capabilities.
Attentive-CoT is an attention-guided fine-tuning objective that improves chain-of-thought performance in multimodal LLMs by delaying answer commitment and increasing sustained visual-token access during rationale generation.
ERGeoBench is a new diagnostic benchmark evaluating MLLMs on four capabilities in three progressive embodied geo-localization settings, finding that models handle high-level semantics but struggle with fine-grained perception and metric localization.
Introduces SANSA paradigm for semantic-agnostic vision-language segmentation via dictionary or example-based prompts, with finetuning delivering up to 20% mIoU gains on the new task while retaining standard performance.
Touch-R1 applies GRPO reinforcement learning on a new 1M tactile dataset and benchmark to train a Qwen2.5-VL-7B model that outperforms baselines on tactile perception and visual-tactile conflict tasks.
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.
SDGBiasBench reveals intrinsic SDG biases in VLMs driven by priors rather than evidence, and CADE mitigates them with up to 25% accuracy gains and 12-point MAE reductions.
WikiVQABench is a human-curated collection of Wikipedia-based VQA items that require both visual evidence and external knowledge from Wikidata to answer correctly.
HalluCXR benchmark shows 61.9-82.3% hallucination rates across VLMs on MIMIC-CXR images, identifies patterns such as length-based risk and over-fabrication of common findings, and demonstrates ensemble mitigation that cuts fabrication by up to 84.8%.
Proposes Spatial Narrative Score (SNS) evaluation for VLMs' camera motion understanding and introduces CaMo model achieving consistent performance on SNS and direct QA.
R3-Streaming uses cascaded control with age-aware memory forgetting and TB-GRPO reinforcement learning to reach SOTA scores of 57.92 on OVO-Bench and 76.36 on StreamingBench with 95-96% fewer visual tokens.
HEED replaces uniform residual alignment with density-weighted alignment using patch self-dissimilarity to improve hybrid VLM distillation, gaining 8.7 points on OCRBench v2 and 5.13 on a 10-benchmark average.
citing papers explorer
-
EgoGapBench: Benchmarking Egocentric Action Selection in Multi-Agent Scenes
EgoGapBench shows humans reliably select egocentric actions in multi-agent scenes while MLLMs systematically choose other agents' actions, and standard egocentric training data fails to close the gap.
-
Learning to Deny: Action Denial in Multimodal Large Language Models
MLLMs drop from over 85% accuracy on action presence to under 50% on matched action-denial videos, exposing a causal verification gap that causal graph prompts partially close.
-
MuseBench: Benchmarking Intent-Level Audiovisual Arts Understanding in MLLMs
MuseBench shows state-of-the-art MLLMs achieve only 48.29% accuracy on intent-level audiovisual arts understanding versus 87.18% for human experts.
-
Bridging VideoQA and Video-Guided Agentic Tasks via Generalized Keyframe Extraction
Introduces VG-GUIBench benchmark and TASKER keyframe extraction algorithm that improves performance on VideoQA and video-guided agentic tasks.
-
FeVOS: Foresight Expression Video Object Segmentation
Introduces the FeVOS task, a 968-clip dataset with foresight expressions, and an MLLM model FeVOS-R1 trained via SFT then RL that reports SOTA on the new task plus generalization to prior RVOS benchmarks.
-
HumanScale: Egocentric Human Video Can Outperform Real-Robot Data for Embodied Pretraining
Processed egocentric human video outperforms teleoperated real-robot trajectories as pretraining data for embodied foundation models, delivering 24% lower validation loss and 52.5-90% higher task success rates under matched post-training protocols.
-
Evaluating and Enhancing Negation Comprehension in Remote Sensing MLLMs
Remote sensing MLLMs perform poorly on negation tasks with hallucinations and accuracy drops, but the NeFo test-time learning method substantially improves negation understanding and generalizes to unseen tasks using ~5% unlabeled test samples.
-
Earth-OneVision: Extending Remote Sensing Multimodal Large Language Models to More Sensor Modalities and Tasks
Earth-OneVision is a unified 2B-parameter RS-MLLM supporting six modalities and nine tasks via FGVLA, SLIS, and PCMA mechanisms plus a 34M QA-pair dataset, reporting competitive or superior benchmark results versus larger models.
-
TVI-CoT: Text-Visual Interleaved Chain-of-Thought Reasoning for Multimodal Understanding
TVI-CoT introduces learnable control tokens <THINK>, <LOOK>, <ANSWER> that let multimodal LLMs interleave textual reasoning with dynamic visual feature access, reporting gains of 3.4-6.1% on eight benchmarks over prior CoT baselines.
-
Closed-Form Spectral Regularization for Multi-Task Model Merging
Iterative solvers in layer-wise model merging act as spectral regularizers on an ill-posed interference operator; closed-form SWUDI and adaptive SWUDI-A match or exceed SOTA merging accuracy with 28-72x wall-clock speedup.
-
SVHighlights: Towards Extremely Long Sport Video Highlight Detection
SVHighlights is the first benchmark for highlight detection in hour-long sports videos, with TF-SELECTOR showing that segment-level LLM scoring outperforms adapted short-video baselines by 2.5-4 points on key metrics.
-
DisasterBench: A Multimodal Benchmark for UAV-Based Disaster Response in Complex Environments
DisasterBench is a new multi-stage multimodal reasoning benchmark for UAV disaster response with 14 scenes and 9 tasks; the accompanying 2B DisasterVL model outperforms open-source MLLMs and approaches GPT-4o efficiency.
-
Eliciting Complex Spatial Reasoning in MLLMs through Wide-Baseline Matching
Authors create ReasonMatch-Bench and DCRL training to boost MLLM performance on wide-baseline matching, reporting gains over baselines while preserving general capabilities.
-
Attention-guided Fine-tuning of Multimodal Large Language Models Improves Chain-of-Thought Reasoning
Attentive-CoT is an attention-guided fine-tuning objective that improves chain-of-thought performance in multimodal LLMs by delaying answer commitment and increasing sustained visual-token access during rationale generation.
-
ERGeoBench:A Comprehensive Benchmark for Embodied Reasoning and Geo-localization in Multimodal Large Language Models
ERGeoBench is a new diagnostic benchmark evaluating MLLMs on four capabilities in three progressive embodied geo-localization settings, finding that models handle high-level semantics but struggle with fine-grained perception and metric localization.
-
Toward Semantic-Agnostic and Shape-Aware Vision-Language Segmentation Models
Introduces SANSA paradigm for semantic-agnostic vision-language segmentation via dictionary or example-based prompts, with finetuning delivering up to 20% mIoU gains on the new task while retaining standard performance.
-
Touch-R1: Reinforcing Touch Reasoning in MLLMs
Touch-R1 applies GRPO reinforcement learning on a new 1M tactile dataset and benchmark to train a Qwen2.5-VL-7B model that outperforms baselines on tactile perception and visual-tactile conflict tasks.
-
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.
-
SDGBiasBench: Benchmarking and Mitigating Vision--Language Models' Biases in Sustainable Development Goals
SDGBiasBench reveals intrinsic SDG biases in VLMs driven by priors rather than evidence, and CADE mitigates them with up to 25% accuracy gains and 12-point MAE reductions.
-
WikiVQABench: A Knowledge-Grounded Visual Question Answering Benchmark from Wikipedia and Wikidata
WikiVQABench is a human-curated collection of Wikipedia-based VQA items that require both visual evidence and external knowledge from Wikidata to answer correctly.
-
HalluCXR: Benchmarking and Mitigating Hallucinations in Medical Vision-Language Models for Chest Radiograph Interpretation
HalluCXR benchmark shows 61.9-82.3% hallucination rates across VLMs on MIMIC-CXR images, identifies patterns such as length-based risk and over-fabrication of common findings, and demonstrates ensemble mitigation that cuts fabrication by up to 84.8%.
-
CaMo: Camera Motion Grounded Evaluation and Training for Vision-Language Models
Proposes Spatial Narrative Score (SNS) evaluation for VLMs' camera motion understanding and introduces CaMo model achieving consistent performance on SNS and direct QA.
-
An Efficient Streaming Video Understanding Framework with Agentic Control
R3-Streaming uses cascaded control with age-aware memory forgetting and TB-GRPO reinforcement learning to reach SOTA scores of 57.92 on OVO-Bench and 76.36 on StreamingBench with 95-96% fewer visual tokens.
-
HEED: Density-Weighted Residual Alignment for Hybrid Vision-Language Model Distillation
HEED replaces uniform residual alignment with density-weighted alignment using patch self-dissimilarity to improve hybrid VLM distillation, gaining 8.7 points on OCRBench v2 and 5.13 on a 10-benchmark average.
-
Towards Unified Surgical Scene Understanding:Bridging Reasoning and Grounding via MLLMs
SurgMLLM unifies high-level reasoning and low-level visual grounding in one MLLM-based model for surgical videos, raising triplet recognition AP from 40.7% to 46.0% on the new CholecT45-Scene dataset with 64,299 annotated frames.
-
Count Anything at Any Granularity
Multi-grained counting is introduced with five granularity levels, supported by the new KubriCount dataset generated via 3D synthesis and editing, and HieraCount model that combines text and visual exemplars for improved accuracy.
-
AnomalyClaw: A Universal Visual Anomaly Detection Agent via Tool-Grounded Refutation
AnomalyClaw turns single-step VLM anomaly judgments into a multi-round tool-grounded refutation process, delivering consistent macro-AUROC gains of 3.5-7.9 percentage points over direct inference across 12 cross-domain datasets.
-
V-ABS: Action-Observer Driven Beam Search for Dynamic Visual Reasoning
V-ABS is an action-observer beam search method with entropy-based adaptive weighting and an 80k-sample SFT dataset that delivers 19.7% average gains on visual reasoning tasks for MLLMs.
-
Beyond GSD-as-Token: Continuous Scale Conditioning for Remote Sensing VLMs
ScaleEarth conditions remote sensing VLMs on continuous GSD via CS-HLoRA and a visual GSD predictor, creating a closed training loop with GeoScale-VQA to achieve SOTA on Earth observation benchmarks.
-
VTAgent: Agentic Keyframe Anchoring for Evidence-Aware Video TextVQA
VTAgent uses a question-guided agent to anchor keyframes for evidence-aware Video TextVQA, delivering up to +12 accuracy and new SOTA results via training-free operation plus SFT and RL.
-
GEASS: Gated Evidence-Adaptive Selective Caption Trust for Vision-Language Models
GEASS is a logit-level gating module that selectively trusts generated captions in VLMs per query by combining clean-path confidence, entropy reduction, and pathway disagreement, improving results on POPE and HallusionBench across four models.
-
Act2See: Emergent Active Visual Perception for Video Reasoning
Act2See trains VLMs via supervised fine-tuning on verified reasoning traces to interleave active frame calls within text CoTs, yielding SOTA results on video reasoning benchmarks.
-
Membership Inference Attacks Against Video Large Language Models
A temperature-perturbed black-box attack infers video training membership in VideoLLMs with 0.68 AUC by exploiting sharper generation behavior on member samples.
-
CGC: Compositional Grounded Contrast for Fine-Grained Multi-Image Understanding
CGC improves fine-grained multi-image understanding in MLLMs by constructing contrastive training instances from existing single-image annotations and adding a rule-based spatial reward, achieving SOTA on MIG-Bench and VLM2-Bench with transfer gains to other multimodal tasks.
-
OptiVerse: A Comprehensive Benchmark towards Optimization Problem Solving
OptiVerse is a new benchmark spanning neglected optimization domains that shows LLMs suffer sharp accuracy drops on hard problems due to modeling and logic errors, with a Dual-View Auditor Agent proposed to improve performance.
-
DistortBench: Benchmarking Vision Language Models on Image Distortion Identification
Vision-language models achieve at most 61.9% accuracy on identifying image distortion types and severities, falling short of human majority-vote performance at 65.7%.
-
DO-Bench: An Attributable Benchmark for Diagnosing Object Hallucination in Vision-Language Models
DO-Bench is a controlled benchmark that attributes VLM object hallucination errors to textual prior pressure, perceptual limits, or their interaction via two diagnostic dimensions and metrics.
-
VisPCO: Visual Token Pruning Configuration Optimization via Budget-Aware Pareto-Frontier Learning for Vision-Language Models
VisPCO uses continuous relaxation, straight-through estimators, and budget-aware Pareto-frontier learning to automatically discover optimal visual token pruning configurations that approximate grid-search results across VLMs and benchmarks.
-
MMR-AD: A Large-Scale Multimodal Dataset for Benchmarking General Anomaly Detection with Multimodal Large Language Models
MMR-AD is a new benchmark dataset showing that current generalist MLLMs lag industrial needs for anomaly detection, with Anomaly-R1 delivering better results through reasoning and RL.
-
AdverMCTS: Combating Pseudo-Correctness in Code Generation via Adversarial Monte Carlo Tree Search
AdverMCTS frames code generation as a minimax game where an attacker evolves tests to expose flaws in solver-generated code, yielding more robust outputs than static-test baselines.
-
AdaSpark: Adaptive Sparsity for Efficient Long-Video Understanding
AdaSpark delivers up to 57% FLOP reduction in Video-LLMs for long videos through adaptive cube- and token-level sparsity without apparent loss in performance on hour-scale benchmarks.
-
Open-Ended Video Game Glitch Detection with Agentic Reasoning and Temporal Grounding
Introduces the first benchmark for open-ended video game glitch detection with temporal localization and proposes GliDe, an agentic framework that achieves stronger performance than vanilla multimodal models.
-
ID-Selection: Importance-Diversity Based Visual Token Selection for Efficient LVLM Inference
ID-Selection combines importance scoring with iterative diversity suppression to prune 97.2% of visual tokens in LVLMs while retaining 91.8% performance and cutting FLOPs by over 97% without retraining.
-
SVAgent: Storyline-Guided Long Video Understanding via Cross-Modal Multi-Agent Collaboration
SVAgent improves long video question answering by constructing storylines via multi-agent collaboration and aligning cross-modal predictions for more robust, human-like reasoning.
-
Focus Matters: Phase-Aware Suppression for Hallucination in Vision-Language Models
Suppressing low-attention tokens during the focus phase of vision-encoder processing reduces object hallucinations in LVLMs while preserving caption quality and adding negligible inference time.
-
V-Reflection: Transforming MLLMs from Passive Observers to Active Interrogators
V-Reflection introduces a think-then-look mechanism where MLLM latent states actively interrogate visual features via two-stage distillation from a box-guided teacher to a dynamic autoregressive student, narrowing the fine-grained perception gap on benchmarks.
-
CFMS: Towards Explainable and Fine-Grained Chinese Multimodal Sarcasm Detection Benchmark
CFMS is the first fine-grained Chinese multimodal sarcasm benchmark with detailed annotations, paired with a PGDS reinforcement learning strategy that improves model results on sarcasm tasks.
-
LongVideo-R1: Smart Navigation for Low-cost Long Video Understanding
LongVideo-R1 trains a reasoning agent on 33K trajectories to intelligently select informative video clips via iterative refinement and RL, achieving better accuracy-efficiency tradeoffs on long video QA benchmarks.
-
ST-BiBench: Benchmarking Multi-Stream Multimodal Coordination in Bimanual Embodied Tasks for MLLMs
ST-BiBench reveals a coordination paradox in which MLLMs show strong high-level strategic reasoning yet fail at fine-grained 16-dimensional bimanual action synthesis and multi-stream fusion.
-
CamReasoner: Reinforcing Camera Movement Understanding via Structured Spatial Reasoning
CamReasoner uses structured O-T-A reasoning and RL on 56k samples to lift camera movement classification from 73.8% to 78.4% and VQA from 60.9% to 74.5% on Qwen2.5-VL-7B.