Metraipolis › archive › dc_d681cba06eec
An Exploratory Ablation of a Small MLA--SSM Hybrid Language Model — arXiv:2609.29618
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.29618
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T04:27:38.512Z
- SHA-256 of the archived text
3fccaa429951d7296330fa5a3fa4f36f72736e9bae46f263a955a630c24367cf- Stable URL of this page
https://metraipolis.com/documento/dc_d681cba06eec
How to cite
An Exploratory Ablation of a Small MLA--SSM Hybrid Language Model — arXiv:2609.29618 — arXiv (en), https://arxiv.org/abs/2609.29618 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d681cba06eec (Metraipolis archive dc_d681cba06eec, sha256 3fccaa429951d7296330fa5a3fa4f36f72736e9bae46f263a955a630c24367cf)
BibTeX
@misc{dcd681cba06eec,
title = {An Exploratory Ablation of a Small MLA--SSM Hybrid Language Model — arXiv:2609.29618},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.29618},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d681cba06eec; sha256 3fccaa429951d7296330fa5a3fa4f36f72736e9bae46f263a955a630c24367cf},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.29618. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We report an exploratory, single-seed ablation of TALH (Adaptive Latent Hybrid), a decoder-only language model with parallel Multi-head Latent Attention (MLA) and a custom recurrent state-space (SSM) branch. Five variants, spanning 117--217M estimated active parameters per token, are trained from scratch on a FineWeb sample for the same number of optimisation steps and tokens. In this specific setup, removing the SSM branch gives the largest degradation in validation perplexity (MLA-only PPL 315), whereas removing MLA has a much smaller effect (SSM-only PPL 239). A dense-FFN hybrid obtains PPL 231, compared with 240 for the tested top-2 ternary-MoE hybrid, while using 3.87 GB less peak training memory. We also preserve a preliminary Apple M3 timing observation: among the five unoptimised implementations, MLA-only has the flattest measured time-to-first-token curve from 512 to 2,048 prompt tokens, although the dense Transformer is much faster in absolute terms. Because the runs are single-seed, parameter counts are unmatched, the evaluation stream may overlap the training source, and raw repeated timing records are unavailable, these results support implementation-specific hypotheses rather than general conclusions about MLA, SSMs, or mixture-of-experts models. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-25 in cs.AI). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.29618 — 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 2026-10-08T03:50:58.110Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).