FlatSounds benchmark shows state-of-the-art V2A models rely more on text captions than visual input for physical and semantic accuracy, with captions improving correctness but degrading temporal alignment.
hub
PhysBench: Benchmarking and enhancing vision-language models for physical world understanding
19 Pith papers cite this work. Polarity classification is still indexing.
19
Pith papers citing it
hub tools
citation-role summary
background 2
dataset 1
citation-polarity summary
representative citing papers
WMW audits VLMs by requiring typed physical state-transition traces and using a verifier to detect inconsistencies missed by answer-only evaluation, with TraceBank as a released resource of synthetic scenarios.
ESI-Bench shows active exploration outperforms passive observation in multimodal LLMs on spatial tasks but reveals failures from poor action choices and overconfident belief commitment unlike humans.
Temporal information in Video-LLMs is encoded well by video-centric encoders but disrupted by standard projectors; time-preserved MLPs plus AoT supervision yield 98.1% accuracy on arrow-of-time and gains on other temporal tasks.
A new benchmark converts video clips into shared grounded event records and tests models across physics, semantic, and control prompts under original, shuffled, ablated, and masked conditions, finding selective robustness and weak spatial performance.
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.
SCP defines a new benchmark task for predicting spatial causal outcomes beyond direct observation and shows that 23 leading models lag far behind humans on it.
SandboxVLM enhances VLMs' spatial intelligence by encoding 3D geometry with abstract bounding boxes in a four-stage zero-shot pipeline, yielding an 8.3% improvement on SAT Real benchmark.
A vision-language framework generates text-based rigid-body scene configurations from videos using motion reasoning and optical flow, reporting 0.30 IoU on CLEVRER (7x over baselines) and transfer to 235 real videos.
GeoWorld-VLM aligns VLM image features with intermediate representations from camera-conditioned world models via fine-tuning only the encoder and projector, yielding ~4% gains on What'sUp and VSR spatial benchmarks across two VLM backbones.
PDI-Bench computes 3D projective residuals from segmented and tracked points to quantify geometric inconsistency in AI-generated videos.
Video-LLMs fail physical reasoning due to semantic prior dominance rather than perception deficits; a new programmatic adversarial curriculum and visual-anchored reasoning chain enable substantial gains via standard LoRA fine-tuning.
Multimodal LLMs significantly underperform humans at spotting objects that break 3D consistency in multi-view image pairs.
Embodied AI requires query-conditioned world models that select the simplest physical abstraction sufficient to answer intervention queries.
PhysBrain 1.0 extracts scene elements, spatial dynamics, actions and depth relations from human egocentric video to create QA supervision for VLMs, then transfers the resulting physical priors to VLA policies via capability-preserving adaptation.
A compact language model trained on scaled synthetic nuclear reactor control data exhibits variance collapse and emergent concentration on a single actuation strategy driven by physical execution success.
MASS adds spatiotemporal motion signals and 3D grounding to VLMs and releases MASS-Bench, yielding physics-reasoning performance within 2% of Gemini-2.5-Flash after reinforcement fine-tuning.
Longitudinal poll data from 471 students in AI courses shows a shift toward preferring human intelligence, reaching 65% in technical courses and 90% in design courses by 2026.
citing papers explorer
No citing papers match the current filters.