3737 papers
This paper formalizes the loss of diversity in classifier-free guidance as "generative distortion," characterizes its emergence via a high-dimensional phase transition using statistical physics, and proposes a novel guidance schedule with a negative-guidance window to mitigate variance shrinkage while preserving class separability.
This paper establishes a rate-distortion theorem demonstrating that hallucinations in large language models are an inevitable consequence of information-theoretic optimal memory compression when storing sparse facts, forcing the model to confidently assign high scores to non-facts rather than abstain.
The paper proposes GVP-WM, a planning method that leverages learned action-conditioned world models to ground zero-shot video-generated plans into dynamically feasible action sequences by optimizing latent trajectories that satisfy physical constraints while preserving semantic alignment with the original video.
This paper challenges the assumption that numerical stability governs generation quality in Decentralized Diffusion Models, demonstrating instead that aligning routing decisions with the experts whose training data best matches the current denoising state is the critical factor for achieving high-quality outputs.
This paper proposes a Contextual Thompson Sampling approach for educational recommender systems that leverages learner data to generate personalized exercise sequences, effectively optimizing skill gain and enabling scalable, adaptive instruction in digital learning environments.
This paper establishes the universality of high-dimensional spectral behavior and statistical limits for asymmetric rank-one spiked tensor models with non-Gaussian noise, demonstrating that the maximum-likelihood estimator's performance matches the Gaussian case under finite fourth-moment assumptions.
The paper introduces BLITZRANK, a principled zero-shot ranking framework that leverages tournament graphs to extract maximal information from -wise comparisons, achieving superior accuracy with significantly reduced token costs compared to existing methods.
This paper introduces Fine-grained Group Policy Optimization (FGO), a reinforcement learning algorithm that effectively compresses verbose Chain-of-Thought reasoning in Large Language Models while simultaneously addressing the data inefficiency and entropy collapse limitations of Group Relative Policy Optimization (GRPO).
The paper proposes GOT-JEPA, a model-predictive pretraining framework that learns to predict robust tracking models from corrupted observations to improve generalization, and introduces OccuSolver to enhance occlusion handling through iterative, object-aware visibility estimation.
The paper proposes LexiSafe, a theoretically grounded offline safe reinforcement learning framework that employs lexicographic prioritization to strictly enforce safety constraints while optimizing task performance, offering improved guarantees and empirical results over existing methods for safety-critical cyber-physical systems.
The paper introduces ZACH-ViT, a compact Vision Transformer that eliminates positional embeddings and the [CLS] token to achieve permutation-invariant processing, demonstrating that this architecture is particularly effective for few-shot medical imaging tasks with weak spatial priors while remaining competitive on datasets with stronger anatomical structures.
This paper introduces new hard benchmarks for Constraint Satisfaction Problems derived from statistical physics to demonstrate that, contrary to some claims, classical heuristics currently outperform Graph Neural Networks on truly difficult instances.
This paper demonstrates that fully autonomous AI analysts can cheaply replicate the analytic diversity and conflicting conclusions observed in human many-analyst studies, revealing that empirical results are highly sensitive to analytic choices and prompting a new transparency norm requiring multiverse-style reporting and full prompt disclosure for AI-generated science.
This paper addresses the challenge of approximating unknown subadditive set functions by developing active querying methods to minimize additive error through the strategic disclosure of subset values, thereby reducing the distance between minimal and maximal completions in both offline and online settings.
This paper investigates how Large Language Models trained on the BabyLM dataset often fail to learn specific grammatical rules because early erroneous biases, driven by misleading bigram statistics, become entrenched and persist throughout training, hindering efficient learning.
The paper introduces CARE, an evidence-grounded agentic framework that enhances clinical accountability and reasoning accuracy in multi-modal medical AI by decomposing tasks into specialized modules for entity proposal, pixel-level localization, and evidence-based reasoning, thereby outperforming state-of-the-art models on medical VQA benchmarks.
This paper introduces CFG-Ctrl, a unified framework that reinterprets Classifier-Free Guidance as a control problem and proposes Sliding Mode Control CFG (SMC-CFG) to overcome the instability and overshooting of linear methods by enforcing nonlinear feedback for improved semantic alignment and robustness across various guidance scales.
This paper presents the first comprehensive evaluation of parameter-efficient fine-tuning (PEFT) for multitask code analysis, demonstrating that a single shared PEFT module can match or surpass full fine-tuning performance while significantly reducing computational and storage costs, provided that tasks are strategically grouped based on factors like complementarity and stability.
This paper proposes Targeted Reasoning Unlearning (TRU), a novel framework that utilizes a reasoning-based unlearning target to guide models in precisely removing specific undesirable knowledge while preserving general capabilities and enhancing robustness against attacks.
MoE-SpAc is an efficient inference framework for Mixture-of-Experts models on heterogeneous edge devices that repurposes speculative decoding as a predictive sensor for memory management, achieving significant throughput improvements through dynamic workload balancing and asynchronous execution.