Source Themes

Anchored Answers: Unravelling Positional Bias in GPT-2's Multiple-Choice Questions

Using principles from Mechanistic Interpretability, we are able to investigate and correct the positional bias in GPT2 models regarding multiple-choice questions.

Ruizhe Li, Yanjun Gao

Learning to Maximize Mutual Information for Chain-of-Thought Distillation

We propose a new training paradigm for Chain-of-Thought distillation with the perspective from information bottleneck.

Xin Chen, Hanxian Huang, Yanjun Gao, Yi Wang, Jishen Zhao, Ke Ding

Learning to Maximize Mutual Information for Chain-of-Thought Distillation

DR. BENCH: Diagnostic Reasoning Benchmark for Clinical Natural Language Processing

We introduced the first Diagnostic Reasoning Benchmark to evaluate clinical natural language processing models’ abilities in clinical diagnostic reasoning.

Yanjun Gao, Dmitriy Dligach, Timothy Miller, John Caskey, Brihat Sharma, Matthew M Churpek, Majid Afshar