Skip to content

How Strongly Should Task State Influence an LLM Agent?

Sep 2026 · arXiv.org · Vol abs/2609.25686 · 0 citations · 86 references
Computer Science

TL;DR

Across three models, two reasoning regimes, and two domains, four findings hold without per-turn reasoning: displaying accurate state is unreliable, an unverified ledger the agent writes itself beats an accurate checklist it is shown, directives help in proportion to the model's obedience, and enforcement needs no obedience but is bounded by the correctness of its state.

Abstract

Long-horizon assigned work requires an LLM agent to track the state of a task: which steps are done, blocked, cancelled, or open to repetition. Agent systems either keep this state as text in the prompt and rely on the model to read that text, or move the state into a module that enforces it, and each system is evaluated as a whole, so no one knows how much reliability comes from the state being shown, told, or enforced. We fix the task rules, the model, and paired episodes and vary how strongly task state reaches the agent: a raw transcript, an exact checklist, per-turn directives from a state machine compiled from the brief and advanced only by execution receipts, or an enforcement gate on that machine that refuses state-violating actions; every episode is scored by exact payload matching against dynamic ground truth. Across three models, two reasoning regimes, and two domains, four findings hold without per-turn reasoning: displaying accurate state is unreliable, an unverified ledger the agent writes itself beats an accurate checklist it is shown, directives help in proportion to the model's obedience, and enforcement needs no obedience but is bounded by the correctness of its state and by the matcher that maps requests to steps; per-turn reasoning at a 235B agent compresses these separations without repairing the text rungs. The same gate, compiled from $\tau^2$-bench's airline policy, raises a 235B agent's pass$^1$ from 0.39 to 0.54 and changes nothing for a 35B agent that rarely violates the policy; on PM-Bench, where acting turns on recognizing a cue rather than on state, showing the record is the best rung--matching or beating both gates and reversing the ledger-over-checklist finding--and enforcing the matcher's judgement drops a 35B agent below its raw transcript. Enforcement pays when failures are state-decidable and frequent, and hurts when the gate's judgement is wrong.

View source

Similar papers

StateTape: Action-Conditioned Evidence Lifecycle Modeling for Long-Horizon Coding Agents

A novel and scalable framework that rewrites a coding agent's context as the repository changes rather than as the context grows, and can achieve a higher resolve rate in all experiments spanned by six coding agents and three edit-heavy benchmarks with little computational overhead.

Zi-Yang Yu, Liang Zhao, Bo-Wen Zhu et al. · 0 citations

Who Holds the Pen? Let Specifications, Not Agents, Sign Off

Experiments on guideline-following and artifact-generation tasks show that specifications can serve not merely as behavioral guidance, but as authority over compliant execution and completion.

Hai-Qing Li, Xin-Yu Ma, Yin-Hao Wu et al. · 0 citations

World State Generator

The World State Generator is a model that writes a plan as checkable states of the world and keeps that plan aligned with the world it runs in, a model that raises end-to-end success for two open models near 30B parameters over prompting and brings to the level of proprietary model.

Sungheon Jeong, Sanggeon Yun, Ryozo Masukawa et al. · 0 citations

SAGE: Symbolic Action-Gating and Editing for LLM Task Planners

SAGE (Symbolic Action-Gating and Editing), a single-LLM planner built from two lightweight mechanisms: a domain-agnostic symbolic gate that blocks precondition-violating actions with typed reasons as a runtime safety monitor, and a local edit that regenerates only the failed sub-goal's suffix, keeping completed and unt...

T. Bui, Jongsul Moon, Youngouk Kim et al. · 0 citations
Preprint Aug 2026

Multi-Agent AI Safety as an Institutional Design Problem

This is the first paper from POLIS, an ongoing research programme studying algorithmic institutions for multi-agent systems, and asks which parts of an AI institution produce safety and how they do it.

X. Abdullah · 1 citation

Related blog posts

MIT News · Artificial Intelligence Nov 17, 2026

Learning the language of machine learning

By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.