Metraipolis › archive › dc_be24526549a7
Protoreasoning in Tiny Transformers — arXiv:2608.04980
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.04980
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T11:15:13.038Z
- SHA-256 of the archived text
c56106d0bce042b942f2a9ab2b826568fe31ac1f7f1b8b7fa89c67021ae8e2a8- Stable URL of this page
https://metraipolis.com/documento/dc_be24526549a7
How to cite
Protoreasoning in Tiny Transformers — arXiv:2608.04980 — arXiv (en), https://arxiv.org/abs/2608.04980 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_be24526549a7 (Metraipolis archive dc_be24526549a7, sha256 c56106d0bce042b942f2a9ab2b826568fe31ac1f7f1b8b7fa89c67021ae8e2a8)
BibTeX
@misc{dcbe24526549a7,
title = {Protoreasoning in Tiny Transformers — arXiv:2608.04980},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.04980},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_be24526549a7; sha256 c56106d0bce042b942f2a9ab2b826568fe31ac1f7f1b8b7fa89c67021ae8e2a8},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.04980. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We show that tiny transformers can profitably employ a simple form of Chain of Thought, which we call protoreasoning, allowing us to study step-by-step reasoning on ~1M-parameter models and opening up opportunities for much more detailed experimentation and analysis than is feasible for larger models. Current Large Language Models exhibit impressive step-by-step reasoning, but we have yet to understand its generality, i.e., when and how LLMs learn genuinely general algorithms rather than "bags of heuristics." Such questions are hard to settle on compute-intensive frontier models trained on opaque data. To work at model scales far below the threshold for natural-language competence, we define reasoning-friendly tasks on Dyck languages (sentences of correctly nested brackets). We find that protoreasoning traces substantially close the out-of-distribution generalization gap, and ablations confirm that the trace's content, not merely its extra tokens, drives the gain. 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); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2608.04980 — 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 2026-10-31T11:02:13.564Z 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).