Metraipolis › archive › dc_72ff74def391

Routing Without Embeddings: Fast And Interpretable Routing With Regular Expressions — arXiv:2609.34326

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.34326
License (as archived): CC0-1.0
Archived: 2026-10-07T22:38:10.189Z
SHA-256 of the archived text: e9fd5c59e1d605237af214041c8e192db907d8cd18023f2e9c9c1ef05419ed2c
Stable URL of this page: https://metraipolis.com/documento/dc_72ff74def391

How to cite

Routing Without Embeddings: Fast And Interpretable Routing With Regular Expressions — arXiv:2609.34326 — arXiv (en), https://arxiv.org/abs/2609.34326 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_72ff74def391 (Metraipolis archive dc_72ff74def391, sha256 e9fd5c59e1d605237af214041c8e192db907d8cd18023f2e9c9c1ef05419ed2c)

BibTeX

@misc{dc72ff74def391,
  title = {Routing Without Embeddings: Fast And Interpretable Routing With Regular Expressions — arXiv:2609.34326},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34326},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_72ff74def391; sha256 e9fd5c59e1d605237af214041c8e192db907d8cd18023f2e9c9c1ef05419ed2c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34326. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large Language Model (LLM) routers commonly rely on neural query embeddings, with larger encoders expected to better capture query intent and difficulty. Yet scaling Qwen2.5 encoders from 0.5B to 72B parameters brings little improvement in routing accuracy (Figure 1b), suggesting that small encoders may already capture the query properties needed for routing. We therefore investigate which properties matter and whether they can be extracted directly from text without a neural encoder. We introduce REGEXROUTE, a pipeline that uses sparse autoencoders (SAEs) to discover interpretable regular-expression (regex) features. Using unlabeled text, an LLM turns descriptions of grouped SAE latents into regex extractors and refines them to match latent activation patterns. These extractors supply numerical features to a lightweight routing head, eliminating neural encoding at inference (Figure 1a). Across four benchmarks, one fixed set of 128 features achieves 76.43% average routing accuracy, comparable to 76.41% for the strongest neural text encoder baseline, with much smaller latency and strong robustness. These findings establish explicit, interpretable text features as a practical basis for designing and understanding LLM routers.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.34326 — 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 2026-10-07T22:14:28.158Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).