AI Research Papers Daily

Emerging Properties in Unified Multimodal Pretraining

SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Training

Visual Agentic Reinforcement Fine-Tuning

Latent Flow Transformer

Reasoning Path Compression: Compressing Generation Trajectories for Efficient LLM Reasoning

Exploring Federated Pruning for Large Language Models

General-Reasoner: Advancing LLM Reasoning Across All Domains

Reasoning Models Better Express Their Confidence

CS-Sum: A Benchmark for Code-Switching Dialogue Summarization and the Limits of Large Language Models

NExT-Search: Rebuilding User Feedback Ecosystem for Generative AI Search

Training-Free Watermarking for Autoregressive Image Generation

Visionary-R1: Mitigating Shortcuts in Visual Reasoning with Reinforcement Learning

VideoEval-Pro: Robust and Realistic Long Video Understanding Evaluation

Fine-tuning Quantized Neural Networks with Zeroth-order Optimization

Think Only When You Need with Large Hybrid-Reasoning Models

Not All Correct Answers Are Equal: Why Your Distillation Source Matters

SSR: Enhancing Depth Perception in Vision-Language Models via Rationale-Guided Spatial Reasoning

Hunyuan-Game: Industrial-grade Intelligent Game Creation Model

Phare: A Safety Probe for Large Language Models

Reward Reasoning Model

MIGRATION-BENCH: Repository-Level Code Migration Benchmark from Java 8

Lessons from Defending Gemini Against Indirect Prompt Injections

Towards eliciting latent knowledge from LLMs with mechanistic interpretability

Warm Up Before You Train: Unlocking General Reasoning in Resource-Constrained Settings

CompeteSMoE -- Statistically Guaranteed Mixture of Experts Training via Competition

Fixing 7,400 Bugs for 1$: Cheap Crash-Site Program Repair

Truth Neurons

Solve-Detect-Verify: Inference-Time Scaling with Flexible Generative Verifier

Tokenization Constraints in LLMs: A Study of Symbolic and Arithmetic Reasoning Limits

Bidirectional LMs are Better Knowledge Memorizers? A Benchmark for Real-world Knowledge Injection

Understanding Gen Alpha Digital Language: Evaluation of LLM Safety Systems for Content Moderation

The Aloe Family Recipe for Open and Specialized Healthcare LLMs