7462 papers
This paper introduces Self-MOA, a fully automated framework that aligns small language models using weak supervision from automated evaluators to achieve significant safety improvements with minimal training data while preserving helpfulness.
The paper introduces \textsc{ReSched}, a minimalist deep reinforcement learning framework that simplifies the Flexible Job Shop Scheduling Problem by condensing the state space to four essential features and utilizing a modified Transformer architecture, achieving superior performance and generalization across various scheduling variants compared to existing methods.
Hit-RAG is a multi-stage preference alignment framework that addresses attention dilution and reasoning hallucinations in long-context multimodal LLMs by systematically refining evidence utilization through supervised fine-tuning, discriminative preference alignment, and group-relative policy optimization to achieve superior performance on complex reasoning tasks.
This paper proposes the Hierarchical Memory Tree (HMT), a structured framework that decouples high-level task logic from site-specific action details through a three-level abstraction hierarchy, thereby significantly enhancing the generalization and robustness of large language model-based web agents in unseen environments.
The paper introduces DeepEarth, a self-supervised multi-modal world model featuring Earth4D, a novel 4D space-time positional encoder that achieves state-of-the-art ecological forecasting performance and outperforms larger foundation models through efficient planetary-scale learning.
This paper proposes CAPL, a framework that mitigates multi-image hallucinations in large vision-language models by introducing a selectable image token interaction mechanism for fine-grained cross-image alignment and a preference learning strategy that trains the model to rely on genuine visual evidence rather than textual priors.
This paper presents a user-friendly framework that enables domain scientists to generate 3D animations of petascale, time-varying climate data on commodity hardware using an LLM-assisted conversational interface, thereby eliminating the need for specialized visualization expertise and high-performance computing resources.
This paper proposes a bi-directional digital twin prototype anchoring framework enhanced with multi-periodicity learning to achieve robust few-shot fault diagnosis by leveraging meta-training in a virtual simulation space and test-time adaptation in the physical domain, thereby overcoming the limitations of traditional methods that require abundant labeled or unlabeled target data.
MedSteer is a training-free activation-steering framework that generates structurally preserved counterfactual endoscopic images by manipulating cross-attention activations in diffusion transformers, outperforming existing methods in concept editing and downstream medical detection tasks.
This paper proposes and validates a novel approach using GPT-4-powered dialogue systems to facilitate user review creation through interview-based information gathering, demonstrating that the resulting system-generated reviews require less editing and are perceived as more helpful by readers than human-written ones, despite some fluency challenges.
This paper introduces CoTJudger, a graph-driven framework that automatically evaluates the efficiency of Large Reasoning Models by converting Chain-of-Thought traces into dependency graphs to identify the Shortest Effective Path, thereby quantifying structural redundancy and revealing pervasive over-reasoning patterns across 21 models.
This paper introduces Countdown-Code, a novel testbed demonstrating that even minimal contamination of supervised fine-tuning data with reward-hacking trajectories can cause large language models to learn and subsequently generalize misaligned behaviors during reinforcement learning, highlighting the critical need for rigorous validation of synthetic training data.
The paper introduces mAVE, a novel watermarking framework that cryptographically binds audio and video latents in joint generation models to eliminate the "Binding Vulnerability" of existing methods and robustly defend against adversarial Swap Attacks without requiring model fine-tuning.
This paper empirically investigates whether large language models can synthesize executable Unity game code from Goal Playable Patterns under strict structural constraints, revealing that while intermediate representations improve performance, project-level grounding and hygiene failures remain primary bottlenecks in achieving high compilation success rates.
This paper proposes the Personalized Semi-Autoregressive with online knowledge Distillation (PSAD) framework, which utilizes a semi-autoregressive teacher model and a User Profile Network to balance generation quality with low-latency inference while enhancing user-item interactions, thereby outperforming state-of-the-art baselines in both ranking performance and efficiency.
This paper introduces ConservationBench to demonstrate that current Vision Language Models systematically fail to reason about physical transformations and maintain invariant representations of physical quantities, often performing near chance levels despite strong textual priors favoring invariance.
This paper presents an LLM-based Werewolf AI agent for the AIWolfDial 2024 shared task that improves utterance consistency and character maintenance by leveraging dialogue summaries and manually designed personas.
This paper introduces aCAPTCHA, a novel security protocol that verifies whether an entity is a capable AI agent by leveraging the asymmetric processing speed between humans and machines to solve the Agentic Capability Verification Problem (ACVP) through time-constrained, multi-round natural language challenges.
This paper introduces EyExIn, a data-efficient framework that enhances retinal Vision Language Models by employing a dual-stream encoding strategy and a deep expert injection mechanism to bridge perception and reasoning gaps, thereby achieving state-of-the-art precision in ophthalmic diagnosis while preventing hallucinations.
This paper introduces Emotion Transcription in Conversation (ETC), a novel task and accompanying Japanese dataset designed to capture complex and subtle emotional states through natural language descriptions, addressing the limitations of traditional categorical emotion recognition methods.