Metraipolis › archive › dc_911bd7ca03a5

Implicit Rule Induction with Test-Time Task Embeddings in ARC-like Tasks — arXiv:2609.21181

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.21181
License (as archived): CC0-1.0
Archived: 2026-10-10T01:01:43.074Z
SHA-256 of the archived text: 17dbb31f1e4a10353e2e85784be2a4d407d8b04d9517a2020326ad5a5ff829a4
Stable URL of this page: https://metraipolis.com/documento/dc_911bd7ca03a5

How to cite

Implicit Rule Induction with Test-Time Task Embeddings in ARC-like Tasks — arXiv:2609.21181 — arXiv (en), https://arxiv.org/abs/2609.21181 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_911bd7ca03a5 (Metraipolis archive dc_911bd7ca03a5, sha256 17dbb31f1e4a10353e2e85784be2a4d407d8b04d9517a2020326ad5a5ff829a4)

BibTeX

@misc{dc911bd7ca03a5,
  title = {Implicit Rule Induction with Test-Time Task Embeddings in ARC-like Tasks — arXiv:2609.21181},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.21181},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_911bd7ca03a5; sha256 17dbb31f1e4a10353e2e85784be2a4d407d8b04d9517a2020326ad5a5ff829a4},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.21181. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

The Abstraction and Reasoning Corpus and related benchmarks evaluate whether AI models can solve novel reasoning tasks, but often leave unclear whether success reflects inference of the intended underlying rule or reliance on shortcuts. We address this gap by studying test-time task embeddings in Vision ARC (VARC), a model in which a pre-trained backbone is complemented by a trainable embedding representing the transformation rule. In the original VARC, test-time training (TTT) is jointly applied to the backbone and task embedding. Here we introduce a novel two-step TTT protocol: first finetune only the task embedding (Embed-TTT), then freeze it and finetune the backbone. Across ARC-AGI-1, ConceptARC, and two controlled datasets with known rules, Embed-TTT consistently yields improved task embeddings, ones that align better with underlying task rules, improve embedding-based retrieval, and enable accurate linear probing of known rules. Qualitatively, Embed-TTT identifies more semantically meaningful relations between test and train tasks on ARC-AGI-1. We also show that optimizing only task embeddings (less than 0.01% of model parameters) already solves a non-trivial fraction of ARC-AGI-1, ConceptARC, and Mini-ARC tasks, while the full two-step pipeline improves final performance. Finally, we show that Embed-TTT recovers the underlying geometric structure of parametric rules and learns compositional capabilities that enable rule-wise interpolation, but not extrapolation. These findings support a clearer separation between rule induction and rule execution in ARC-like evaluations, motivating benchmarks that better distinguish in-distribution from out-of-distribution rules.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-21 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.21181 — 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 2026-10-10T00:23:42.435Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-21?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).