12021 papers
This paper introduces VLMQ, a post-training quantization framework tailored for vision-language models that leverages a gradient-driven importance factor to address visual over-representation and modality gaps, thereby achieving state-of-the-art performance across various model sizes and low-bit settings.
The paper proposes SGDFuse, a novel two-stage conditional diffusion model guided by Segment Anything Model (SAM) semantic masks, which achieves high-fidelity infrared and visible image fusion by leveraging explicit semantic priors to preserve key targets and minimize artifacts for superior downstream task performance.
This paper proposes a best-first search algorithm utilizing delayed partial expansions to explicitly treat control parameters as decision points within infinite domains, offering a complete and competitive alternative to existing constraint-based approaches for automated planning.
This paper introduces "Answer-Then-Check," a novel safety alignment method that enhances LLM robustness against jailbreak attacks by training models to generate direct answers internally and then critically evaluate their safety before responding, achieving superior protection with reduced over-refusal while maintaining general reasoning capabilities through the newly constructed 80K-sample ReSA dataset.
This paper addresses the inconsistency and structural biases in existing latency metrics for simultaneous speech-to-text translation by introducing a comprehensive meta-evaluation, proposing new metrics (YAAL and LongYAAL) and a resegmentation tool (SoftSegmenter), and implementing these solutions within the OmniSTEval toolkit to enable more reliable system assessments.
The paper introduces LikePhys, a training-free evaluation method using likelihood preferences to assess intuitive physics understanding in video diffusion models, demonstrating that current models show improving capabilities in physical reasoning as they scale despite challenges with complex dynamics.
Phys2Real is a real-to-sim-to-real reinforcement learning framework that enhances sim-to-real transfer for precise robotic manipulation by fusing vision-language model-inferred physical parameter priors with online interactive adaptation through uncertainty-aware ensemble estimation.
CanvasMAR enhances masked autoregressive video prediction by introducing a global "canvas" prior and a motion-aware curriculum to generate high-fidelity, coherent videos with fewer sampling steps, achieving performance that rivals advanced diffusion-based methods.
This paper introduces "Just-In-Time Objectives," a framework that passively observes user behavior to infer and rapidly optimize for specific, real-time goals, enabling large language models to generate specialized tools and responses that significantly outperform standard generic interactions.
The paper introduces 3DThinker, a novel framework that enables vision-language models to perform 3D spatial reasoning from limited views by aligning their internal representations with a 3D foundation model and refining the reasoning process through outcome-based optimization, all without requiring explicit 3D prior inputs or labeled 3D training data.
This study demonstrates that ChatGPT-based coding of communication data performs consistently across gender and racial/ethnic subgroups, matching human rater reliability and validating its potential for large-scale collaborative assessments.
This paper introduces the Collaborative Battleship task to evaluate language models' information-seeking abilities and proposes Bayesian Experimental Design-inspired Monte Carlo inference strategies that significantly enhance both question-asking and answer-accuracy, enabling weaker models to outperform humans and frontier models in strategic decision-making tasks.
This paper introduces REx86, a locally deployable, fine-tuned Qwen2.5-Coder-7B model that significantly enhances x86 assembly reverse engineering by improving code comprehension and accuracy while addressing the privacy and security limitations of cloud-based LLMs.
LA-MARRVEL is a knowledge-grounded, language-aware LLM framework that significantly improves rare disease gene prioritization accuracy by using structured, phenotype-rich prompts to generate clinically robust, ACMG-aligned reasoning without disrupting existing diagnostic pipelines.
This paper introduces the Cultural Reference Transformation (CRT) metric to evaluate how diffusion models navigate the tension between memorization and generalization in culturally iconic contexts, revealing that model behavior depends on distinct recognition and realization mechanisms influenced by factors like data frequency, textual uniqueness, and reference popularity.
This paper introduces Soft Q-based Diffusion Finetuning (SQDF), a novel KL-regularized reinforcement learning method that employs a reparameterized policy gradient of a training-free soft Q-function, enhanced by discount factors, consistency models, and off-policy replay buffers, to effectively align diffusion models with downstream objectives while mitigating reward over-optimization and preserving sample diversity.
This paper introduces XR-DT, an Extended Reality-enhanced Digital Twin framework that integrates a novel Human-Aware Model Predictive Path Integral (HA-MPPI) controller with an attention-based trajectory prediction model to enable safe, efficient, and interpretable motion planning for mobile robots operating alongside humans.
This paper proposes a novel training framework that leverages the -divergence family to explicitly filter incorrect answers and control the precision-diversity trade-off, thereby overcoming the diversity loss inherent in standard Reinforcement Learning and achieving state-of-the-art performance on the Lean theorem-proving benchmark.
This paper proposes a point cloud-based framework for event-driven human pose estimation that leverages spatiotemporal properties through novel temporal slicing and sequencing modules alongside an edge-enhanced representation, achieving improved accuracy and efficiency on the DHP19 dataset without converting event streams into dense frames.
The paper introduces A-3PO, a method that accelerates asynchronous LLM training by 1.8x by approximating the computationally expensive proximal policy in Decoupled PPO through simple interpolation, thereby eliminating the need for extra forward passes while maintaining comparable performance.