Sparse autoencoders inserted into VLMs and trained only for reconstruction can reliably detect adversarial attacks on images, including unseen domains and attack types.
Title resolution pending
19 Pith papers cite this work. Polarity classification is still indexing.
19
Pith papers citing it
verdicts
UNVERDICTED 19representative citing papers
CCTVBench exposes a large gap between standard QA accuracy and contrastive consistency in traffic video reasoning for multimodal LLMs and introduces C-TCD to narrow that gap.
Introduces the UCSF-PDGM-VQA dataset of 2387 QA pairs from 473 glioma MRI studies and demonstrates that state-of-the-art VLMs exhibit modality collapse on multi-sequence 3D medical images.
BRITE benchmark reveals that leading T2V models handle static object composition well but degrade sharply on object-action binding and audio-visual synchronization for implausible prompts.
MMBench-Live introduces an automated multi-agent pipeline and distribution-consistent update strategy to create a continuously evolving multimodal benchmark with 5.9K new instances at low cost.
PGT generates synthetic tasks via geometric overlays on images to supply dense visual supervision, improving spatial and relational understanding in MLLMs by up to 20% on targeted benchmarks.
DMN achieves over 90% attack success rate on GPT-4o, Gemini-2.5-pro and Claude Sonnet 4 by distributing instructions, supplying multimodal evidence, and adding number chain tasks across multiple images.
OProver-32B achieves top Pass@32 scores on MiniF2F, ProverBench, and PutnamBench by combining continued pretraining with iterative agentic proving, retrieval, SFT on repairs, and RL on unresolved cases using a 6.86M-proof dataset.
PA-BDM adapts block diffusion by switching to causal intra-block denoising and dynamically committing reliable prefixes to KV cache, yielding higher accuracy and 71.6% higher throughput than a comparable baseline on document benchmarks.
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.
F^3A is a training-free visual token pruning router that treats pruning as task-conditioned evidence search and allocates a fixed vision token budget using question cues and frozen sparse heads without extra LLM passes.
Probabilistic programs of thought let LLMs produce many program variants from one generation by building a compact probabilistic representation of the token distribution.
A new keyframe selection framework combines structural, tracking, and semantic criteria to select reliable anchor frames for diffusion-based video editing under occlusion.
GAP aligns visual latent reasoning in MLLMs via PCA-mapped decoder outputs, auxiliary visual supervision, and selective capacity-guided training, yielding top supervised performance on a 7B model with evidence that latents carry task-relevant signal.
PRISM interleaves VLM perception and LLM reasoning via a dynamic goal-oriented question-answer pipeline to produce sharper scene descriptions, outperforming prior image-based models on ALFWorld and Room-to-Room.
VANGUARD is a staged-training VLM framework that reports 94% ROC-AUC and 84% F1 on UCF-Crime while adding chain-of-thought reasoning and spatial grounding to video anomaly detection.
NVIDIA releases the Nemotron 3 model family with hybrid Mamba-Transformer architecture, LatentMoE, NVFP4 training, MTP layers, and multi-environment RL post-training for reasoning and agentic tasks.
Proposes Modality-Aware Credit Assignment (MoCA) with blindfolded-reasoning proxy to reward perception fidelity separately from reasoning in VLMs.
Proposes unifying pix and word tokens in generative LMs via per-pixel embeddings, color folding, and unsupervised pretraining, claiming strong small-model performance on limited data.
citing papers explorer
No citing papers match the current filters.