Paper

JobBench: Aligning Agent Work With Human Will

Unreadunread

AI summary

arXiv:2605. 26329v1 Announce Type: new Abstract: Current benchmarks for occupational AI agents are scoped primarily by economic values, telling a replacement story.

Reading status

Priority

About this paper

Authors: Yuetai Li, Yichen Feng, Zhangchen Xu, Zixian Ma, Kaiyuan Zheng, Fengqing Jiang, Xinghua Sun, Rulin Shao, Zichen Chen, Yue Huang, Xinyang Han, Brian Lee, Kayla Xu, Shenglai Zeng, Hang Hua, Xiangliang Zhang, Basel Alomair, Ranjay Krishna, Luke Zettlemoyer, Pang Wei Koh, Bhaskar Ramasubramanian, Luyao Niu, Xiang Yue, Radha Poovendran
Released: 2026-05-27 (2mo ago)
Topic: Related to current project
Relation to my work: Read next because JobBench: Aligning Agent Work With Human Will overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, under, eval, chain, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv: 2605.26329

Ask Claude about this paper

Posts a comment with @claude so the runner spins up a Claude Code session scoped to JobBench: Aligning Agent Work With Human Will. Replies appear in the conversation below.

Comments

No comments yet.

Related records

No links yet.