Metraipolis › archive › dc_9baee8360d44

Sparse Token Routing in Efficient Transformers — arXiv:2608.20632

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.20632
License (as archived)
CC0-1.0
Archived
2026-10-18T10:18:20.803Z
SHA-256 of the archived text
5f3dd8d2c09ec980c0f5c80eec6f5b8a45f9b39d12130cecf2142bebc162128d
Stable URL of this page
https://metraipolis.com/documento/dc_9baee8360d44

How to cite

Sparse Token Routing in Efficient Transformers — arXiv:2608.20632 — arXiv (en), https://arxiv.org/abs/2608.20632 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9baee8360d44 (Metraipolis archive dc_9baee8360d44, sha256 5f3dd8d2c09ec980c0f5c80eec6f5b8a45f9b39d12130cecf2142bebc162128d)
BibTeX
@misc{dc9baee8360d44,
  title = {Sparse Token Routing in Efficient Transformers — arXiv:2608.20632},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.20632},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9baee8360d44; sha256 5f3dd8d2c09ec980c0f5c80eec6f5b8a45f9b39d12130cecf2142bebc162128d},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.20632. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Efficient-transformer research often motivates token pruning and adaptive computation with the claim that not all tokens require equal computational effort. We test this claim end to end using SEWN, a two-stream Transformer that routes tokens through either lightweight or full-capacity processing using a learned gate. Across our experiments, routing introduces negligible accuracy change relative to parameter-matched baselines, while the gate's token-importance signal depends critically on how it is learned. A static lexicon-seeded prior fails a counterfactual faithfulness test on BoolQ, whereas a fully contextual gate achieves highly significant separation ($p<10^{-10}$) on both evaluated tasks without changing task accuracy.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-24 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.20632 — 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 2026-10-18T10:17:37.577Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).