---
id: "2026-10-25-ai-assisted-results-october-25-listing"
url: "https://postcutoff.com/e/2026-10-25-ai-assisted-results-october-25-listing/"
as_of: "2026-10-25T23:49:00+01:00"
date: "2026-10-25"
date_precision: day
category: science
importance: 2
confidence: medium
status: [Awaiting review]
verification: Preprint only
sources: 3
editor: human editor
human_review: null
version: "2026-10-25"
---

As of: 2026-10-25 23:49 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/e/2026-10-25-ai-assisted-results-october-25-listing/

# AI-assisted results in the 25 Oct 2026 sweep

Full title: AI-assisted results in the 25 Oct 2026 sweep: GPT-5.5 Thinking steers a neutral-atom tweezer array, GBClaw grain-boundary agent, ChatGPT Astra-built bone-marrow model

Catalogue of AI-disclosing preprints from the 17–20 Oct 2026 arXiv listings that no earlier roundup cited, read on 25 Oct. An LLM (GPT-5.5 Thinking via ChatGPT Plus) chose the control strategy for a 10×10 acousto-optic tweezer array, cutting non-uniformity from 20.30% to 1.57%. GBClaw (Georgia Tech, GPT-5.6 Sol and Luna) ran grain-boundary simulations in nickel. A Bristol modeller built a dynamic bone-marrow model through dialogue with ChatGPT (Astra, Extra High).

## Key facts
- LLM steering of acousto-optic tweezer arrays (Cheng Sheng, …, Gao-xiang Li; Huazhong Normal University and CAS Wuhan; 2610.25014): 'We test this role on the optical control subsystem of a neutral-atom apparatus using OpenAI GPT-5.5 Thinking through ChatGPT Plus.' The LLM recommended the model, update scale and next action. Deterministic software did the fitting, and an experimentalist authorized every hardware update. When separable row–column correction stalled, the LLM moved the workflow to a drift-corrected local response-matrix method. The Gaussian-peak coefficient of variation fell from 20.30% to 1.57% in 30 rounds
- GBClaw (Yazhuo Liu, Kunqing Ding, …, Ting Zhu; Georgia Tech, Osaka, USC; 2610.28629): agentic framework (orchestrator, specialized agents, LAMMPS/OVITO/ASE tools, hierarchical memory) that 'used both GPT-5.6 Sol and GPT-5.6 Luna, with task-dependent model routing'. Applied to 14 general asymmetric tilt grain boundaries in nickel, it characterized interfacial defect structures and deformation modes and found that secondary GB misfit dislocations strongly change shear strength. Researcher-guided follow-up calculations traced an anomalously high shear resistance to metastable boundary structures
- Diabetic bone-marrow dynamic model (Paolo Madeddu, University of Bristol; 2610.24728): 'AI assistance was provided through iterative dialogue with ChatGPT (OpenAI; Astra, Extra High reasoning setting). ChatGPT assisted evidence extraction, proposed regulatory rules and alternative assumptions, and supported coding and simulation.' The three-level logical model ranks endothelial ROS suppression as the best single intervention (score 24.2); ROS suppression combined with OSM blockade or p66Shc loss scores 37.9. The author says experiments must test this
- Checked, no AI disclosure found in the PDF: symmetric-group generation by prefix reversals (2610.30892), Wilf-equivalences / Dumont permutations conjecture (2610.30975), polynomial coefficients (2610.29132), Toeplitz semi-commutators (2610.30787), 2D MHD stability (2610.28308), radiating-gas model (2610.29462). Associated primes of monomial ideals (2610.27916) declares no generative AI use

## What happened

Sunday 25 Oct had no new arXiv announcement. The sweep listed 17–22 Oct papers that earlier roundups had not cited. Most of
them only mention LLMs as a topic (benchmarks, applications). PDFs of the uncited maths papers and of the three papers that
gave an AI model a working role were scanned (pdftotext). Quotes above are from the PDFs.

## Why it matters

The tweezer paper is a small but concrete case: a general-purpose chat model chose the experimental strategy on real quantum
hardware, while a human authorized each step. None of the items is a famous-problem result.

## What is disputed or not yet verified

- Verification: Unrefereed arXiv preprints; the tweezer result is a physical experiment

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 178 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 117 days after its cutoff
- Gemini 4 Argon (training cutoff March 2026): 208 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 147 days after its cutoff

## Sources

1. [arXiv 2610.25014: Large Language Model Steering of Acousto-Optic Tweezer Arrays](https://arxiv.org/abs/2610.25014) (arxiv.org, paper)
2. [arXiv 2610.28629: GBClaw, an agentic AI framework for atomistic grain boundary research](https://arxiv.org/abs/2610.28629) (arxiv.org, paper)
3. [arXiv 2610.24728: Evidence-informed dynamic modelling and intervention prioritisation in diabetic bone marrow](https://arxiv.org/abs/2610.24728) (arxiv.org, paper)

## Changes

- 2026-10-25 (filed): Created

## Related

- 2026-10-24: [oai-math-revised: mathematicians launch a community hub to revise OpenAI's AI-written math manuscripts (guest post on Tao's blog)](https://postcutoff.com/e/2026-10-24-oai-math-revised-community-hub/index.md)
- 2026-10-24: [AI-assisted results in the 24 Oct 2026 sweep](https://postcutoff.com/e/2026-10-24-ai-assisted-results-october-24-listing/index.md)