Metraipolis › archive › dc_b344259f5643

Joint or Per-Option? Compact Encoders for Closed-Set Decisions from 17 Million to 1 Billion Parameters — arXiv:2610.30970

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.30970
License (as archived)
CC0-1.0
Archived
2026-10-25T08:36:24.043Z
SHA-256 of the archived text
b8f7ac78a5bcc5fb3d454c2bc45d168cf2dccfce10ad3662c9410d9b83e43bdf
Stable URL of this page
https://metraipolis.com/documento/dc_b344259f5643

How to cite

Joint or Per-Option? Compact Encoders for Closed-Set Decisions from 17 Million to 1 Billion Parameters — arXiv:2610.30970 — arXiv (en), https://arxiv.org/abs/2610.30970 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b344259f5643 (Metraipolis archive dc_b344259f5643, sha256 b8f7ac78a5bcc5fb3d454c2bc45d168cf2dccfce10ad3662c9410d9b83e43bdf)
BibTeX
@misc{dcb344259f5643,
  title = {Joint or Per-Option? Compact Encoders for Closed-Set Decisions from 17 Million to 1 Billion Parameters — arXiv:2610.30970},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.30970},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b344259f5643; sha256 b8f7ac78a5bcc5fb3d454c2bc45d168cf2dccfce10ad3662c9410d9b83e43bdf},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.30970. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Many language-model pipelines end in a closed-set decision, such as selecting a tool. Prompting a large language model for it is slow and order-sensitive. We ask how small an encoder for such decisions can be and whether it should read the options jointly or separately. In a three-seed study at 32 million parameters, per-option scoring beats joint scoring on a held-out sample of a 37-benchmark decision suite by 5.8 points of chance-corrected skill at matched optimizer steps and by 2.8 at approximately matched training compute; two external benchmarks agree, though not significantly on their unseen tasks. Crossing reader with initialisation shows that the advantage requires a reranker initialisation, whose query-candidate pair input matches per-option reading; from the raw encoder the two readers are indistinguishable. Per-option scoring needs 5 to 13 times more encoder passes: at most twice the median latency, but more than four times the mean. Of six released joint encoders (17 million to 1 billion parameters), the 68- and 150-million-parameter models exceed a 435-million-parameter DeBERTa-v3 zero-shot classifier by 4.9 and 5.9 accuracy points on 209 overlap-audited System1-Arena tasks and are at least on par with it on the 145 tasks unseen in training. The smallest answers a median request in 14 milliseconds on 16 processor threads, against 728 for the classifier. Reordering options changes 12 to 23% of the joint models' answers (per-option scoring: none). Up to 68 million parameters, they lead their size range on a public leaderboard's private, mainly in-distribution tests.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.30970 — 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 2026-10-25T08:31:27.778Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).