Metraipolis › archive › dc_d14b493eb4ee

Visualizing Graph-to-Answer Mechanism Recovery in Materials-Science Hypothesis Generation — arXiv:2608.04170

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.04170
License (as archived)
CC0-1.0
Archived
2026-10-31T11:13:36.489Z
SHA-256 of the archived text
2defc096c93ba26f520d9b6eb73e3147968ad610c2a8d1c3f445e0d79c5ebaa9
Stable URL of this page
https://metraipolis.com/documento/dc_d14b493eb4ee

How to cite

Visualizing Graph-to-Answer Mechanism Recovery in Materials-Science Hypothesis Generation — arXiv:2608.04170 — arXiv (en), https://arxiv.org/abs/2608.04170 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d14b493eb4ee (Metraipolis archive dc_d14b493eb4ee, sha256 2defc096c93ba26f520d9b6eb73e3147968ad610c2a8d1c3f445e0d79c5ebaa9)
BibTeX
@misc{dcd14b493eb4ee,
  title = {Visualizing Graph-to-Answer Mechanism Recovery in Materials-Science Hypothesis Generation — arXiv:2608.04170},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.04170},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d14b493eb4ee; sha256 2defc096c93ba26f520d9b6eb73e3147968ad610c2a8d1c3f445e0d79c5ebaa9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.04170. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

AI co-scientists can generate fluent materials-science hypotheses, but fluency does not show that an answer preserves a scientifically meaningful mechanism. We present a graph-to-answer mechanism-tracing case study for Graph-PRefLexOR-8B, a Qwen3-8B model adapted to expose distinct stages for brainstorming, graph construction, pattern extraction, and synthesis. We organize semantic backtracking, graph corruption, activation-based recovery measurements, and layer-by-token-region grids into a visual diagnostic workflow for inspecting this pathway. Across 100 open-ended materials-science questions, final answers remain closest to the model's own structured stages, especially synthesis. Under graph corruption, a full sweep over 37 residual-stream checkpoints, the embedding output and 36 transformer blocks, shows little mechanism recovery in the earlier transition region at layers 7--10, recovery instead concentrates in late synthesis and answer-start regions around layers 30 and 36. The workflow is intended to help scientists and model developers identify where a generated hypothesis loses or regains mechanism support before it is passed to downstream experimental planning.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-06 in cs.AI).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.04170 — 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 2026-10-31T11:02:13.487Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-06?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).