AI
RefusalGuard: Geometry-Preserving Fine-Tuning for Safety in LLMs
Researchers have found that fine-tuning language models for specific tasks can cause them to lose their ability to refuse harmful requests. This is a ...
AI
AI
Preference-Based Self-Distillation: Beyond KL Matching via Reward Regularization
Researchers have proposed a new method for on-policy distillation called Preference-Based Self-Distillation (PBSD). Unlike previous methods that rely ...
AI
GRALIS: Fusing Coalition and Gradient Attribution with Closed-Form Conservation Error and Finite-Sample Guarantees
Researchers have developed GRALIS, a new method for explaining how deep neural networks make decisions. GRALIS combines two existing methods, coalitio...
AI
S-AI-Recursive: Convergent Recursive Reasoning
Researchers have introduced a new AI architecture called S-AI-Recursive, which uses a bio-inspired approach to implement reasoning as a closed-loop it...
AI
Component-Aware Structure-Preserving Style Transfer for Satellite Visual Sim2Real Data Construction
Researchers have developed a method for creating synthetic images of satellites that closely resemble real-world images while preserving important ann...
AI
Behavior-Consistent Deep Reinforcement Learning
Researchers have proposed a new approach to reinforcement learning called Q-Value Expectile Disagreement (QED). The goal of QED is to reduce the varia...
AI
Lost in Sampling: Assessing Lexical Reachability in LLMs via the Word Coverage Score (WCS)
Researchers have developed a metric called the Word Coverage Score (WCS) to measure how well large language models (LLMs) can generate diverse and uni...
AI
Chatterbox-Flash: Prior-Calibrated Block Diffusion for Streaming Zero-Shot TTS
Researchers have developed Chatterbox-Flash, a text-to-speech model that can generate speech without prior training data. The model uses a block-diffu...
AI
Audio Interaction Model
Researchers have proposed an Audio Interaction Model that allows for continuous and interactive audio processing. Unlike current large language models...
AI
INFUSER: Influence-Guided Self-Evolution Improves Reasoning
Researchers have developed a new framework called INFUSER that improves the reasoning abilities of language models through self-evolution. The framewo...
AI
Valid Inference with Synthetic Data via Task Exchangeability
Researchers Lezhi Tan and Tijana Zrnic propose a new approach to using synthetic data in scientific research. They argue that synthetic data can be bi...
AI
Crypto x AI, AI x Crypto: A Survey
A survey paper explores the intersection of cryptocurrency and artificial intelligence (AI). The authors systematize existing work, summarize key take...