Tech Meridian ← ENTITY INDEX
PROMY MERIDIAN RU

TOPIC · ENTITY #5494

multimodal large language models (MLLMs)

Related event timeline, sources and context from the news index.

EVENT TIMELINE

5

RESEARCH · 1 SOURCE · arXiv cs.AI

CST-Bench: benchmark and annotations for counterfactual spatial reasoning, plus CT-OPSD method

The authors introduce CST-Bench, a dataset of 7,082 examples for counterfactual spatial reasoning covering 6 intervention types and 12 transformation operators with dense causal annotations (intervention sources, affected/invariant relations, expected effects, metric changes). They benchmark general-purpose multimodal LLMs on CST-Bench, compare post-training strategies, and propose CT-OPSD (Causal Transition On-Policy Self-Distillation) to teach counterfactual state transitions via annotated teacher chains distilled into a student model, reporting CST-Bench as a challenging evaluation and reusable supervision resource.

6.0

RESEARCH · 1 SOURCE · arXiv cs.AI

Brain-aligned evaluation of multimodal LLMs for social and emotional understanding of cartoons

This arXiv paper compares diffusion-based and autoregressive multimodal LLMs to continuous human ratings of Theory of Mind and affective valence across three short cartoon movies, assessing temporal Pearson correlations, distributional differences, and the effect of preceding scene context. The authors also used model and human ratings in subject-specific regressions to predict regional fMRI responses, finding that MLLMs capture aspects of human temporal variation (stronger for valence than ToM), that context helps in some cases, and that model-to-brain alignment is model- and stimulus-dependent with limited overlap between model and human encodings.

6.0

RESEARCH · 1 SOURCE · arXiv cs.AI

GeoSpatialBench: benchmark for geospatial reasoning and constrained spatial editing (arXiv v1)

The paper (arXiv:2610.12881v1) introduces GeoSpatialBench, a new benchmark evaluating spatial cognition and constrained spatial editing across real-world urban GeoJSON and map-image modalities, covering 28 topological, directional, metric, and ordering relations. Evaluation of frontier multimodal LLMs shows GeoJSON representations outperform map images, multimodal inputs give no consistent gain, metric relations and constrained edits are notably weaker than QA, and many failed edits violate specified geometric constraints.

6.0

RESEARCH · 1 SOURCE · arXiv cs.AI

DEEPO: dual-entropy enhanced policy optimization to reduce hallucination in MLLMs

The paper (arXiv:2609.28570v1) proposes DEEPO, a dual-stage RL enhancement for multimodal LLMs that combines semantic-entropy-triggered expert prefixes to restore advantage variance and advantage-sign-aware Renyi preconditioning to counteract logit saturation. Both components individually outperform GRPO and together yield statistically significant gains on the VideoMMMU benchmark (+4.0, 95% CI [1.1, 6.9]); DEEPO reduces hallucination while preserving accuracy and training stability.

6.0

RESEARCH · 1 SOURCE · arXiv cs.AI

PolyBridgeBench: benchmarking MLLMs for physics-grounded bridge design

PolyBridgeBench is an executable benchmark that asks multimodal LLMs to produce complete node–member–material bridge topologies from a visual scene and engineering constraints, then runs deterministic legality checks and a native dynamic physics simulation. The benchmark reports deterministic validity, dynamic functional success, and post-failure recovery under a fixed interaction budget; experiments on six representative MLLMs across 189 levels reveal a large gap between deterministic validity and dynamic success, strong sensitivity to material budgets, and limited ability to repair after failures.

6.0