JEPA · 2026-10-01

OverForge: Reasoning Through Strategies and Tactics Helps Cooperative Lifelong Adaptation

Oana Madalina Fron, Ojas Shirekar, Chirag Raman

arXiv:2609.39727PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

OverForge is a training-free hierarchical architecture for cooperative language-model agents. It separates strategic reasoning over roles/divisions of labour from tactical reasoning over actions inside each agent’s private, partner-conditioned world model, coupled by a metacognitive Prefrontal Cortex Module that imagines consequences and commits when confident. In OvercookedV2, it delivers 7 soups in a connected kitchen vs 3 for flat LLM baselines.

Why it matters

The authors address coordination that requires both persistent strategies and moment-to-moment tactics, especially under unfamiliar partners and across episodes. By placing a JEPA-inspired predictive hierarchy inside each agent’s world model (rather than using a single flat planner), the approach aims to guide tactical adaptation with stable coordination strategies and support cross-episode partner knowledge for continual adaptation.

Method

  • OverForge uses a training-free hierarchical controller that separates strategy (roles/divisions of labour) from tactics (state-dependent primitive actions) within each agent’s private, partner-conditioned world model.
  • A metacognitive Prefrontal Cortex Module couples levels by generating strategy-action branches, valuing first-action options via pairwise LLM judges, valuing rollouts via partner-conditioned imagined states, and committing when confidence passes a threshold.
  • Branching samples strategies (persistent) and executable first actions (tactical), then rolls forward partner-conditioned imagined transitions while re-proposing tactics at each depth; confidence gates when to stop deliberation.
Abstract (from arXiv)

Cooperative language-model agents must coordinate over long horizons and adapt to changing environments and to partners with unfamiliar conventions, yet existing agents map observations to actions without separating persistent coordination strategies from their tactical execution. We introduce OverForge, a training-free hierarchical architecture that separates strategic reasoning over roles and divisions of labour from tactical reasoning over actions within each agent's private, partner-conditioned world model. A metacognitive Prefrontal Cortex Module couples the two levels by forming strategy-action branches, imagining their consequences with a forward model, and committing when confident. In OvercookedV2, OverForge delivers 7 soups in a connected kitchen versus 3 for each flat LLM baseline, retains agreed roles, and adopts roles proposed by unfamiliar partners. Ablations and a fixed-strategy probe show that persistent strategies guide tactical adaptation while each reasoning level contributes to coordination. Memory restarts show that cross-episode partner knowledge supports task performance and partner prediction, linking the hierarchy to continual adaptation.

Related papers

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-09-30 · Indexed by Pub-AI: 2026-10-01 UTC

preprint

OverForge: Reasoning Through Strategies and Tactics Helps Cooperative Lifelong Adaptation

Oana Madalina Fron, Ojas Shirekar, Chirag Raman

Preprint posted: 2026-09-30 · Original record ↗

arxiv
2609.39727
Explicit source versions (1)
  • v1 · 2026-09-30
  1. Preprint available

    Source date: 2026-09-30 · First detected: 2026-10-25 UTC · Historical baseline

    Source evidence (1)
Abstract · arxiv

Cooperative language-model agents must coordinate over long horizons and adapt to changing environments and to partners with unfamiliar conventions, yet existing agents map observations to actions without separating persistent coordination strategies from their tactical execution. We introduce OverForge, a training-free hierarchical architecture that separates strategic reasoning over roles and divisions of labour from tactical reasoning over actions within each agent's private, partner-conditioned world model. A metacognitive Prefrontal Cortex Module couples the two levels by forming strategy-action branches, imagining their consequences with a forward model, and committing when confident. In OvercookedV2, OverForge delivers 7 soups in a connected kitchen versus 3 for each flat LLM baseline, retains agreed roles, and adopts roles proposed by unfamiliar partners. Ablations and a fixed-strategy probe show that persistent strategies guide tactical adaptation while each reasoning level contributes to coordination. Memory restarts show that cross-episode partner knowledge supports task performance and partner prediction, linking the hierarchy to continual adaptation.