The Order Matters: Sequential Fine-Tuning of LLaMA for Coherent Automated Essay Scoring (arxiv.org)

1h fine-tuningllama

PADD: Path-Aligned Decompression Distillation for Non-Router Teacher to Guide MoE Student Learning (arxiv.org)

1h moe

Do Vision-Language Models See or Guess? Measuring and Reducing Textual-Prior Reliance with a Phrasing-Controlled Benchmark (arxiv.org)

1h

WebChallenger: A Reliable and Efficient Generalist Web Agent (arxiv.org)

1h

Prefilling-dLLM: Predictive Prefilling for Long-Context Inference in Diffusion Language Models (arxiv.org)

1h

Small Data, Big Noise: Adversarial Training for Robust Parameter-Efficient Fine-Tuning (arxiv.org)

1h fine-tuning

Multilingual Word-Level Forced Alignment with Self-Supervised Representations and Learned Dynamic Programming (arxiv.org)

1h

REAL: A Reasoning-Enhanced Graph Framework for Long-Term Memory Management of LLMs (arxiv.org)

1h

Continual LLM Upcycling: A Predictor-Gated Bank-Wise Sparsity Training Recipe for Dense-to-Sparse LLMs (arxiv.org)

1h

Pushing the Limits of LLM Tool Calling via Experiential Knowledge Integration and Activation (arxiv.org)

1h

Density Field State Space Models: 1-Bit Distillation, Efficient Inference, and Knowledge Organization in Mamba-2 (arxiv.org)

1h

Measuring Human Value Expression in Social Media Texts: Calibrated LLM Annotation and Encoder Transfer (arxiv.org)

1h

Does Reasoning Preserve Alignment? On the Trustworthiness of Large Reasoning Models (arxiv.org)

1h

Attention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix It (arxiv.org)

1h fine-tuning

VISTA: A Versatile Interactive User Simulation Toolkit for Agent Evaluation (arxiv.org)

1h

Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models (arxiv.org)

1h

SpeechJBB: Probing Safety Alignment and Comprehension in Large Audio Language Models under Code-Switched Speech (arxiv.org)

1h

Mechanistic Analysis of Alignment Algorithms in Language Models (arxiv.org)

1h

Streaming Knowledge Compilation: Proactive Materiality-Scored Pinning for Time-Evolving LLM Wikis (arxiv.org)

1h

Benchmarking and Exploring the Capabilities of LLMs for Attack Investigations (arxiv.org)

1h

Enhancing Multilingual LLM-based ASR with Mixture of Experts and Dynamic Downsampling (arxiv.org)

1h mixture-of-experts

SpenseGPT: Practical One-shot Pruning Enabling Sparse and Dense GEMMs for LLM Inference (arxiv.org)

1h

How Does Reasoning Flow? Tracing Attention-Induced Information Flow for Targeted RL in LLMs (arxiv.org)

1h

RedAct: Redacting Agent Capability Traces for Procedural Skill Protection (arxiv.org)

1h

Training LLMs to Enforce Multi-Level Instruction Hierarchies via Gravity-Weighted Direct Preference Optimization (arxiv.org)

1h

Data Journalist Agent: Transforming Data into Verifiable Multimodal Stories (arxiv.org)

1h

What Really Matters for Table LLMs? A Meta-Evaluation of Model and Data Effects (arxiv.org)

1h

Automated Alignment between Elicitation Interviews and Requirements (arxiv.org)

1h

Mitigating hallucinations in healthcare LLMs with granular fact-checking and domain-specific adaptation (arxiv.org)

1h

inversedMixup: Data Augmentation via Inverting Mixed Embeddings (arxiv.org)

1h