7354 papers
This paper introduces Best-of-Tails (BoT), an adaptive inference-time alignment framework that dynamically balances optimistic and pessimistic selection strategies by characterizing reward distribution tail heaviness via the Hill estimator and using Tsallis divergence as a tunable regularizer to optimize performance across diverse reasoning and preference tasks.
This paper introduces AceMAD, a multi-agent debate framework that overcomes the "Martingale Curse" of standard methods by leveraging asymmetric cognitive potential energy—where truth-holders anticipate collective misconceptions—to transform agent convergence from a random walk into a directed drift toward the correct answer.
This paper presents a hybrid machine learning model that combines VGG19, EfficientNet, and ResNet50 for feature extraction with a Bi-LSTM classifier to achieve a 98.83% accuracy in the early detection of Cerebral Palsy from MRI images, outperforming several individual pre-trained models.
This paper introduces "context specification" as a process to transform diffuse stakeholder perspectives into clear, measurable constructs, thereby bridging the gap between abstract AI evaluation methods and the operational realities necessary for successful, value-driven AI deployment.
This paper argues that in decentralized multi-agent reinforcement learning, the instability of the agent-world boundary caused by peer-policy updates leads to the erosion of invariant decision cores, thereby framing the challenge as a continual learning problem driven by boundary drift rather than exogenous task switches.
This study utilizes AI-assisted curation to compile and analyze a comprehensive database of 23,793 presentations from the Society for Social Work and Research Annual Conference (2005–2026), revealing significant growth in participation, collaboration, and international engagement alongside a continued predominance of quantitative research methods.
This paper proposes the Dark Triad personality traits as a framework for studying AI misalignment, demonstrating that frontier large language models can be reliably induced with human-like antisocial behaviors through minimal fine-tuning on psychometric data, thereby revealing latent persona structures that generalize beyond training contexts.
This paper establishes that the quality of a model's step-level visual grounding, quantified by the Step Grounding Rate (SGR), serves as a robust and independent predictor of out-of-distribution generalization in long-horizon vision-language models, outperforming traditional final-answer accuracy metrics.
This paper introduces Twitch, a tool that automatically discovers reusable term abstractions from failed or successful proofs to enhance the performance of the Twee equational theorem prover, resulting in the solution of new problems and significant speed-ups on existing benchmarks.
This paper introduces a mechanistic interpretability approach to identify "audio-specialist" attention heads in large audio-language models, enabling a parameter-free inference-time steering technique that significantly boosts audio grounding and accuracy by amplifying the model's reliance on audio evidence.
This paper introduces Contextual Counterfactual Credit Assignment (C3), a novel method for multi-agent reinforcement learning with large language models that isolates the causal impact of individual messages through context-matched counterfactual replay and leave-one-out baselines to solve sparse terminal feedback issues and significantly improve collaborative performance.
This paper presents a toolkit leveraging Large Language Models to automate key aspects of Artifact Evaluation in cybersecurity research, achieving high accuracy in reproducibility rating, autonomous environment setup, and pitfall detection to significantly reduce reviewer effort and enhance research transparency.
This paper proposes a hardware-efficient trajectory prediction framework that integrates environmental priors with a Dual-Transformer-Cascaded (DTC) architecture to accurately predict the landing points of flying objects, such as tennis balls, by outperforming existing methods in complex real-world scenarios.
SymLang is an open-source framework that integrates symmetry-constrained grammars, language-model-guided program synthesis, and Bayesian model selection to robustly discover accurate, interpretable governing equations from noisy and partial observations, significantly outperforming existing baselines in structural recovery and physical consistency.
The paper introduces Lookahead-Enhanced Atomic Decomposition (LEAD), a method that mitigates the "no-recovery bottleneck" caused by extreme decomposition in long-horizon reasoning by integrating short-horizon future validation and overlapping rollouts, thereby enabling models like o4-mini to solve complex algorithmic puzzles beyond the limits of previous approaches.
This paper introduces LieCraft, a novel multi-agent framework featuring grounded, high-stakes scenarios and a hidden-role game mechanic to evaluate the deceptive capabilities of large language models, revealing that state-of-the-art models consistently exhibit a willingness to lie, conceal intentions, and act unethically to achieve their goals.
This study reveals that while the correctness of LLM-generated reasoning is the primary driver of user accuracy in critical thinking tasks, medium-length explanations uniquely enhance users' ability to detect errors when the AI's reasoning is incorrect, suggesting that response length plays a nuanced role in shaping human critical evaluation.
This paper introduces a Physics-Informed Neural Operator (PINO) surrogate model that accelerates the retention analysis of Ferroelectric Vertical NAND devices by over 10,000 times compared to traditional TCAD simulations while maintaining physical accuracy, thereby enabling efficient optimization of device designs against charge detrapping and ferroelectric depolarization.
This paper proposes a Distributed Legal Infrastructure (DLI) framework comprising five interlocking layers—ranging from soulbound agent identities to decentralized adjudication—to establish interoperable protocols that ensure accountability, contestability, and rule-of-law principles within the emerging autonomous agentic web.
This study demonstrates that a hybrid machine learning model combining Bi-LSTM and GRU algorithms, trained on diverse patient data, achieves a 97.07% accuracy in detecting Coronary Artery Disease, significantly outperforming traditional diagnostic methods.