Metraipolis › archive › dc_ec2eed2eca35

Long-Horizon State Tracking in LLMs: Executing MD5 through a Deep Sequence of Dependent Tool Calls — arXiv:2609.00012

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.00012
License (as archived): CC0-1.0
Archived: 2026-10-17T06:30:48.752Z
SHA-256 of the archived text: 430114b96ae714fee53ae8602b4b2390fc57b04b0c850c902a7951d5bd23c641
Stable URL of this page: https://metraipolis.com/documento/dc_ec2eed2eca35

How to cite

Long-Horizon State Tracking in LLMs: Executing MD5 through a Deep Sequence of Dependent Tool Calls — arXiv:2609.00012 — arXiv (en), https://arxiv.org/abs/2609.00012 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ec2eed2eca35 (Metraipolis archive dc_ec2eed2eca35, sha256 430114b96ae714fee53ae8602b4b2390fc57b04b0c850c902a7951d5bd23c641)

BibTeX

@misc{dcec2eed2eca35,
  title = {Long-Horizon State Tracking in LLMs: Executing MD5 through a Deep Sequence of Dependent Tool Calls — arXiv:2609.00012},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.00012},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ec2eed2eca35; sha256 430114b96ae714fee53ae8602b4b2390fc57b04b0c850c902a7951d5bd23c641},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.00012. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Long-horizon tasks remain uncommon in large language model (LLM) evaluation, and for a reason: when each step depends on the last, per-step accuracy that looks excellent in isolation decays catastrophically, as errors cascade and the end-to-end failure probability grows sharply with length. Existing agentic benchmarks report end-to-end success but confound this state-tracking difficulty with instruction interpretation, give no control group that isolates it, and are vulnerable to shortcuts such as a hallucinated final answer, so they cannot say why a long run fails. Whether an LLM can carry exact intermediate state across many tool calls at all is itself not well established. We test this cleanly by having the model compute a cryptographic hash, MD5, step by step: a sequence of $196$ dependent tool calls over $64$ rounds while it carries four $32$-bit words $(a,b,c,d)$ in its own context from one call to the next. Interpretation is trivial and, because we implement MD5 from scratch (RFC~1321), we align every call to the ground-truth trace and check the digest to the bit, so any failure is pure bookkeeping. gpt-oss-120b, a mixture-of-experts model with only $\sim$5.5B active parameters per token, at temperature $0$ with a short fixed prompt, carries the full state across all $196$ calls and returns the correct digest on a majority of completed runs. In the strongest setting we replace every primitive tool with a second LLM, so a driver and a worker compute the whole hash from scratch with no exact-arithmetic oracle in the loop. Two ingredients decide success and neither changes the weights: keeping the model's own reasoning in its context each turn, and voting over a thinking-enabled worker to remove its modular-arithmetic slips. We localize the residual failures by origin, separating state-carrying from arithmetic and from serving.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-02 in cs.MA).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Cryptography and Security (cs.CR); Multiagent Systems (cs.MA).
Pagina del lavoro: https://arxiv.org/abs/2609.00012 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T06:14:52.154Z dalla pagina https://arxiv.org/catchup/cs.MA/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).