13 papers
Not All LLM Reasoning is Visible in the Chain-of-Thought
Vatsal Baherwani, Tom Goldstein, Ashwinee Panda
A key question for AI safety is whether a language model expresses all of its reasoning in its output tokens. We demonstrate a concrete failure mode where frontier models exhibit i…
Multi-Token Prediction via Self-Distillation
John Kirchenbauer, Abhimanyu Hans, Brian Bartoldson +3
Existing techniques for accelerating language model inference, such as speculative decoding, require training auxiliary speculator models and building and deploying complex inferen…
Speculating Experts Accelerates Inference for Mixture-of-Experts
Vivan Madan, Prajwal Singhania, Abhinav Bhatele +2
Mixture-of-Experts (MoE) models have gained popularity as a means of scaling the capacity of large language models (LLMs) while maintaining sparse activations and reduced per-token…
From Refusal Tokens to Refusal Control: Discovering and Steering Category-Specific Refusal Directions
Rishab Alagharu, Ishneet Sukhvinder Singh, Shaibi Shamsudeen +2
Language models are commonly fine-tuned for safety alignment to refuse harmful prompts. One approach fine-tunes them to generate categorical refusal tokens that distinguish differe…
Identifying and Evaluating Inactive Heads in Pretrained LLMs
Pedro Sandoval-Segura, Xijun Wang, Ashwinee Panda +4
Attention is foundational to large language models (LLMs), enabling different heads to have diverse focus on relevant input tokens. However, learned behaviors like attention sinks,…
FineGRAIN: Evaluating Failure Modes of Text-to-Image Models with Vision Language Model Judges
Kevin David Hayes, Micah Goldblum, Vikash Sehwag +3
Text-to-image (T2I) models are capable of generating visually impressive images, yet they often fail to accurately capture specific attributes in user prompts, such as the correct…