arXiv:2611.25211v1 Announce Type: new Abstract: Long event streams require identifying each change and applying the changes in order. We train a selector on a frozen language model to read each transition. With API-only access, we ask the model to list the transitions. An external integer register composes them in order. A confidence rule can flag uncertain updates or retain several candidates. We study products of permutations of five items, a task that fixed-depth, log-precision transformers cannot solve on all input lengths under the standard conjecture that $\mathrm{TC}^0 \ne \mathrm{NC}^1$. On two-event windows, a 27B model answered 3 of 100 correctly without visible reasoning and 96 of 100 when it generated intermediate steps. A frozen 0.8B model with a trained selector read every event in 299 of 299 windows of 43,687 events each. The same package passed 298 of 299 new windows on a fresh machine, missing the target of 0.99 for the one-sided 95% lower confidence bound. A separate follow-up sized to allow one failure yielded 472 of 473. On 1,000-event room-swap documents, three of ten models solved all five documents by reasoning alone. All seven models tested both ways at this length solved all five with extraction and the register, with no chain of thought requested. With concurrent extraction, median wall-clock time fell from 136.1 to 10.4 s for Sonnet 5.5 and from 313.5 to 28.0 s for GPT-6.1 Sol. At fixed per-token rates, mean cost per document fell by 55% for Sonnet 5.5 and 44% for GPT-6.1 Sol. A regular expression with an executor also solved all 15 fixed-format documents. On 500 free-form issue-tracker comments, a separately adapted 4B reader identified the recorded value in 68.8% of cases, compared with 61.0% for a frequency baseline fitted on the same training projects.
Learning to Read Events for Exact State Tracking with Frozen Language Models
A frozen 0.8B language model paired with a trained selector and an external integer register read every event in 299 of 299 windows of 43,687 events each, according to arXiv paper 2611.25211v1. The same package passed 298 of 299 new windows on a fresh machine, missing the 0.99 target for the one-sided 95% lower confidence bound, while a follow-up sized to allow one failure yielded 472 of 473. On 1,000-event room-swap documents, all seven models tested both ways solved all five documents with extraction and the register and no chain of thought, and concurrent extraction cut median wall-clock time from 136.1 to 10.4 s for Sonnet 5.5 and from 313.5 to 28.0 s for GPT-6.1 Sol, with mean cost per document falling 55% and 44% respectively at fixed per-token rates.
Offload ordered state updates to an external register and have the model only extract each transition; the source solved all five 1,000-event documents this way and cut Sonnet 5.5 median latency from 136.1 to 10.4 s with concurrent extraction.
more in playbook: Designing agent systemsRun your AI side-project on zahid.host
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.