1 citations · 1 across the 7 of their papers we have counts for
9 papers
Generalization Dynamics of LM Pre-training
Jiaxin Wen, Zhengxuan Wu, Dawn Song +1
People typically assume that LMs stably mature from pattern-matching parrots to generalizable intelligence during pre-training. We build a toy eval suite and show this mental model…
Automated Researchers Can Mitigate Well-characterized Alignment Failures
Chen Yueh-Han, Jiaxin Wen, Jan Hendrik Kirchner
Automating alignment research may accelerate progress toward aligned AI, but whether it does is hard to measure. Luckily, many alignment failures, such as deception, sycophancy, an…
Single Canonical Prompts Underestimate LLM Safety's Surface-Form Sensitivity
Yongxi Zhou, Junwei Yao, Yuanzhe Liu +4
A benchmark score is a measurement instrument, yet most benchmarks read each item at a single canonical surface form. We ask whether that reading is faithful: when an item's intent…
How Much Can Repeated Running Buy? A Measured Variance Ceiling for LLM Code Evaluation
Yongxi Zhou, Lai Yun Choi, Jiaxi Wen +2
A benchmark score is a measurement whose uncertainty has two sources: an aleatoric component from stochastic decoding and an epistemic component from the finite problem set. Using…
Efficient Hyperparameter Tuning via Trajectory Invariance Principle
Bingrui Li, Jiaxin Wen, Zhanpeng Zhou +2
As hyperparameter tuning becomes increasingly costly at scale, efficient tuning methods are essential. Yet principles for guiding hyperparameter tuning remain limited. In this work…
Measuring Harmfulness of Computer-Using Agents
Aaron Xuxiang Tian, Ruofan Zhang, Janet Tang +3
Computer-using agents (CUAs), which can autonomously control computers to perform multi-step actions, might pose significant safety risks if misused. However, existing benchmarks m…
Unsupervised Elicitation of Language Models
Jiaxin Wen, Zachary Ankner, Arushi Somani +10
To steer pretrained language models for downstream tasks, today's post-training paradigm relies on humans to specify desired behaviors. However, for models with superhuman capabili…
Predicting Empirical AI Research Outcomes with Language Models
Jiaxin Wen, Chenglei Si, Yueh-han Chen +2
Many promising-looking ideas in AI research fail to deliver, but their validation takes substantial human labor and compute. Predicting an idea's chance of success is thus crucial…
Adaptive Deployment of Untrusted LLMs Reduces Distributed Threats
Jiaxin Wen, Vivek Hebbar, Caleb Larson +9
As large language models (LLMs) become increasingly capable, it is prudent to assess whether safety measures remain effective even if LLMs intentionally try to bypass them. Previou…