4 papers
QVal: Cheaply Evaluating Dense Supervision Signals for Long-Horizon LLM Agents
Sergio Hernández-Gutiérrez, Matteo Merler, Ilze Amanda Auzina +3
LLM agents increasingly act over long horizons, where a single trajectory can contain hundreds or thousands of actions. In these settings, outcome-only rewards provide too sparse g…
RevengeBench: Reverse Engineering Code-Space Policies from Behavioral Experiments
Babak Rahmani, Sebastian Dziadzio, Joschka Strüber +2
For most of scientific history, researchers studying behavior could only infer hidden mechanisms from outward actions: an inverse problem that becomes more tractable when observati…
Intrinsic Credit Assignment for Long Horizon Interaction
Ilze Amanda Auzina, Joschka Strüber, Sergio Hernández-Gutiérrez +3
How can we train agents to navigate uncertainty over long horizons? In this work, we propose ÎBelief-RL, which leverages a language model's own intrinsic beliefs to reward interme…
Recursive Decomposition with Dependencies for Generic Divide-and-Conquer Reasoning
Sergio Hernández-Gutiérrez, Minttu Alakuijala, Alexander V. Nikitin +1
Reasoning tasks are crucial in many domains, especially in science and engineering. Although large language models (LLMs) have made progress in reasoning tasks using techniques suc…