19 papers
GrAInS: Gradient-based Attribution for Inference-Time Steering of LLMs and VLMs
Duy Nguyen, Archiki Prasad, Elias Stengel-Eskin +1
Inference-time steering methods offer a lightweight alternative to fine-tuning large language models (LLMs) and vision-language models (VLMs) by modifying internal activations at t…
Multi-Attribute Steering of Language Models via Targeted Intervention
Duy Nguyen, Archiki Prasad, Elias Stengel-Eskin +1
Inference-time intervention (ITI) has emerged as a promising method for steering large language model (LLM) behavior in a particular direction (e.g., improving helpfulness) by inte…
A History-Aware Visually Grounded Critic for Computer Use Agents
Jaewoo Lee, Zaid Khan, Archiki Prasad +7
Various test-time interventions for Computer Use Agents (CUAs), including critic models, have been developed to improve performance through pre-execution action evaluation in compl…
PRInTS: Reward Modeling for Long-Horizon Information Seeking
Jaewoo Lee, Archiki Prasad, Justin Chih-Yao Chen +3
Information-seeking is a core capability for AI agents, requiring them to gather and reason over tool-generated information across long trajectories. However, such multi-step infor…
Effective Reasoning Chains Reduce Intrinsic Dimensionality
Archiki Prasad, Mandar Joshi, Kenton Lee +2
Chain-of-thought (CoT) reasoning and its variants have substantially improved the performance of language models on complex reasoning tasks, yet the precise mechanisms by which dif…
AVSD: Adaptive-View Self-Distillation by Balancing Consensus and Teacher-Specific Privileged Signals
Duy Nguyen, Hanqi Xiao, Archiki Prasad +7
Self-distillation enables language models to learn on-policy from their own trajectories by using the same model as both student and teacher, with the teacher being conditioned on…