Pratinav Seth
3 indexed papers
Publications per year
Top categories
Frequent co-authors
Research Timeline
This paper investigates the effectiveness of filtering and recovery methods in synthetic post-training pipelines.
The paper presents ALIGNBEAM, a training-free method to improve the safety of large language models by translating anchor logits into the target model's vocabulary and selecting the safest continuation.
The paper presents AuditKIT, a library for evaluating language models and agents by recording various measurements in one result representation.
Papers
AuditKIT: A Unified Evaluation Spine for Benchmarking, Judging, and Comparing Large Language Models and Agentic Systems
Shreeyans Arora, Utsav Avaiya, Zera Lyngkhoi, Hem Gosalia +3 more
The paper presents AuditKIT, a library for evaluating language models and agents by recording various measurements in one result representation.