LLM Digest

Story

arxiv_cs_cl · Oct 26, 2026 · paper

Source brief

SEMA-Auto: Sighted and Efficient Multi-Turn Attack with Autonomy

arxiv.orgOct 26, 2026
original source linked

In brief

Existing jailbreak attackers fall into two classes: static planners fix their single- or multi-turn adversarial plans in advance, remaining simple but losing sight of the victim's replies, wasting turns and limiting t...

Feed lens

eval

Read the original at arxiv.org →Open in live feed Read that day’s brief

SEMA-Auto: Sighted and Efficient Multi-Turn Attack with Autonomy

Earlier in this thread 4 items

Dependency-Aware Trajectory Refinement for Efficient Multi-Turn Agent Fine-Tuning

CoSA: The Conflict of Skills Attack in Real-World LLM Agent Ecosystems

Multi-Turn Reinforcement Learning on Amazon SageMaker AI: A hands-on guide

One Word Opens the Gate: The Option-Channel Attack on Typed Decision Models as Agent Guardrails