Hong Chen

icon location I am a graduate student of cyberspace security in the University of the Chinese Academy of Sciences

Results 1 comments of Hong Chen

Code Generation Evals should parse code from LM response

> From personal experience, generating code in the instruct format (prompt used in [this paper](https://openreview.net/forum?id=y0GJXRungR)) works very well for most (API) models. On some internal evals on fresh leetcode style...