agent/papers/items/2026-2606-22673-agentlens-interpretable-safety-steering-via-mechanistic-subspaces-for-multi-turn.md

Paper: AgentLens: Interpretable Safety Steering via Mechanistic Subspaces for Multi-Turn Coding Agent

---

type: paper title: "AgentLens: Interpretable Safety Steering via Mechanistic Subspaces for Multi-Turn Coding Agent" authors: Weidi Luo, Qiming Zhang, Yihao Quan, Mingyu Jin, Jie Cai, Chaowei Xiao, Jingcheng Niu, Zhen Xiang year: 2026 venue: arXiv url: https://arxiv.org/abs/2606.22673 code_url: source: arxiv collected_at: 2026-07-08 published_at: 2026-06-21 updated_at: 2026-06-21 status: queued relevance: high topics:

• agent-evaluation
• agent-safety
• coding-agent
• tool-use methods:

benchmarks:

models:

datasets:

• cs.AI
• cs.SE related_concepts:

related_jobs:

related_experiments:

related_projects:

collection_score: 18 collection_queries: agent-safety, coding-agent

One-line Takeaway

Auto-collected from arXiv because it matched the Agent collection queries. Needs human skim.

Why Collected

• matched queries: agent-safety, coding-agent
• inferred topics: agent-evaluation, agent-safety, coding-agent, tool-use
• arXiv categories: cs.AI, cs.SE
• collection score: 18

Review Checklist

• Does this paper directly inform Agent architecture, evaluation, memory, tools, safety, coding agents, GUI/browser agents, or multi-agent workflows?
• Does it include a benchmark, dataset, code, or reproducible experimental setup?
• Should it be promoted from queued to skimmed or summarized?

Links

• arXiv: https://arxiv.org/abs/2606.22673