REVEAL: Reverse Engineering Agentic Latents from Frontier-Model Traces

arXiv:2610.16288v2 Announce Type: new Abstract: Frontier LLM agents perform complex interactive tasks, but their inference costs motivate transferring their experience to smaller open-source agents. Distillation and latent communication typically require target-model training or access to source-model internals. We study how textual frontier-model trajectories can provide reusable local controls for a frozen agent. Inspired by protocol reverse engineering, we introduce extit{Agentic Latents}, comprising anchored states and dynamic profiles of feedback--response cycles elicited by replay in the target model. Our training-free framework, extbf{REVEAL} (extbf{REV}erse extbf{E}ngineering extbf{A}gentic extbf{L}atents), uses these profiles and trajectory outcomes to match and select local events from successful and failed rollouts of the same task and source model. Their response-transition contrasts yield a target-specific steering direction, applied before each response. Within-benchmark evaluations on Terminal-Bench 2.1 and Humanity's Last Exam with tools report higher accuracy than unmodified agents across three targets. Representation analyses and behavioral case studies support reusing frontier-model experience as local activation control.

Related

• From Interaction Traces to Persistent Skills: Online Evolution for Computer-Use Agents
• SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories
• Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills

Source: arXiv cs.AI | 2026-10-15