llm-kb
← Back to research
Paper

First systematic study of chat-to-agent transfer of additive activation steering

Researchers present the first systematic study of how additive activation steering—calibrated in single-turn chat—transfers to tool-using ReAct agents. Using a matched-information design, they compare behavioral and representational effects of steering vectors in plain chat versus ReAct episodes, controlling for KV-cache contamination.

0 engagement·1 source·Fri, Jul 10, 2026, 07:21 AM
The paper introduces a matched-information design where the same items are rendered as plain chat or as a ReAct tool-use episode, with matched-norm random-direction controls. The transcript is re-encoded every turn to exclude KV-cache contamination. This work addresses the gap between steering calibration (done in chat) and deployment (in agents).

Entities

Additive activation steering(concept)ReAct agents(concept)

Related

Paper

LLM/VLA models enable prompt-driven exploration in RL

A new research paper proposes using large language models (LLMs) and vision-language-action (VLA) models to drive exploration in reinforcement learning by modifying natural language prompts, which induce global behavioral changes beyond standard action noise. This approach could help policies escape weak local optima more effectively.

0 engagement·1 source·arxiv
Thu, Jul 9, 2026, 06:00 PM
arXiv
Paper

KV-PRM: Efficient Process Reward Modeling via KV-Cache Transfer for Multi-Agent Test-Time Scaling

A new paper introduces KV-PRM, a process reward model that reuses KV-cache from the base model to avoid re-encoding entire trajectories, reducing scoring cost from quadratic to linear in sequence length. This enables efficient test-time scaling for long-context multi-agent systems.

0 engagement·1 source·arxiv
Fri, Jul 10, 2026, 07:16 AM
arXiv
Paper

Researchers propose memory agent to combat behavioral state decay in long-horizon tasks

A new arXiv paper identifies 'behavioral state decay' as a failure mode where decision-relevant information gets buried in long trajectories. The authors propose a separate memory agent that actively updates a structured memory bank alongside an unmodified action agent, rather than relying on passive retrieval.

0 engagement·1 source·arxiv
Thu, Jul 9, 2026, 05:26 PM
arXiv
Paper

GATS framework eliminates LLM calls during agent planning inference

Researchers propose GATS (Graph-Augmented Tree Search), a planning framework that uses a layered world model and UCB1-based tree search to avoid LLM inference during planning, reducing computational cost and stochasticity. The approach outperforms LATS and ReAct on multi-step planning tasks.

0 engagement·1 source·arxiv
Thu, Jul 9, 2026, 07:34 PM
arXiv
Paper

CLAP paper proposes direct VLM-to-VLA adaptation to isolate semantic transfer

A new arXiv paper introduces CLAP, a method that converts pretrained vision-language models (VLMs) into vision-language-action models (VLAs) with minimal architectural changes, avoiding the distribution mismatch caused by predicting actions as bare numeric tokens. The approach aims to clarify how VLM capabilities transfer across model scales for robot control.

0 engagement·1 source·arxiv
Thu, Jul 9, 2026, 10:34 PM
arXiv