Metraipolis › archive › dc_a88f142d3f04
Programs-of-Layers in LLMs through the Lens of Cortical Areas — arXiv:2609.31360
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.31360
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T01:38:41.977Z
- SHA-256 of the archived text
e494676379f4543bca316bd70db15e02d6bb89e2b16433ccac0e2baf4e1e601a- Stable URL of this page
https://metraipolis.com/documento/dc_a88f142d3f04
How to cite
Programs-of-Layers in LLMs through the Lens of Cortical Areas — arXiv:2609.31360 — arXiv (en), https://arxiv.org/abs/2609.31360 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a88f142d3f04 (Metraipolis archive dc_a88f142d3f04, sha256 e494676379f4543bca316bd70db15e02d6bb89e2b16433ccac0e2baf4e1e601a)
BibTeX
@misc{dca88f142d3f04,
title = {Programs-of-Layers in LLMs through the Lens of Cortical Areas — arXiv:2609.31360},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.31360},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a88f142d3f04; sha256 e494676379f4543bca316bd70db15e02d6bb89e2b16433ccac0e2baf4e1e601a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.31360. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Inference in LLMs is conventionally a fixed-depth, fixed-order forward pass through every layer, regardless of how difficult the input is. The human brain does not work this way: using the thalamus as a central hub, it routes information flexibly to all regions of the cortex according to demand. Li et al. (2026) recently showed, with a system they call program-of-layers (PoLar), that transformers can be given an analogous flexibility if their layers are treated as a library of functions rather than a fixed sequence. Performance improves over the standard forward pass when each input is dynamically routed through an adaptive sequence of skipped or repeated contiguous layer blocks. We reconstructed PoLar's diagnostic MCTS in more detail than the original paper and applied it across 5 models. We reproduced several of PoLar's findings: skipping outperformed the standard pass, repeating outperformed skipping, and combining both outperformed either alone. Shorter programs sufficed for easier questions, while harder questions required more layer repeats. However, we failed to replicate the main claim regarding their learned router for single-shot inference: its top-ranked prediction consistently collapsed back to the standard pass, even though its top-k predicted programs, taken together, did show a real accuracy gain. Beyond reproduction, we find that a small number of generic programs are enough to solve most of the questions. We also provide a much deeper analysis of these programs' structure and robustness: for example, we found that programs that correct errors are highly brittle: undoing even a single edit inside a program typically breaks the correction. Connecting this to the brain's routing mechanisms, PoLar mirrors principles of thalamo-cortical coordination between cortical-area-like transformer layers. We publicly release the code at this https URL Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.31360 — 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 2026-10-08T01:12:14.375Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).