Metraipolis › archive › dc_f451964a8409

DARTS: Decoder-Aware Representation Tuning via Surgery for Model Merging — arXiv:2608.28547

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.28547
License (as archived)
CC0-1.0
Archived
2026-10-17T11:06:39.594Z
SHA-256 of the archived text
addb7448fe244ff9a737f3367fac2d9f2ff4912b19f7f0485360bc320b34fe76
Stable URL of this page
https://metraipolis.com/documento/dc_f451964a8409

How to cite

DARTS: Decoder-Aware Representation Tuning via Surgery for Model Merging — arXiv:2608.28547 — arXiv (en), https://arxiv.org/abs/2608.28547 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f451964a8409 (Metraipolis archive dc_f451964a8409, sha256 addb7448fe244ff9a737f3367fac2d9f2ff4912b19f7f0485360bc320b34fe76)
BibTeX
@misc{dcf451964a8409,
  title = {DARTS: Decoder-Aware Representation Tuning via Surgery for Model Merging — arXiv:2608.28547},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.28547},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f451964a8409; sha256 addb7448fe244ff9a737f3367fac2d9f2ff4912b19f7f0485360bc320b34fe76},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.28547. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Model merging combines multiple task-specific fine-tuned LLMs into a single multi-task model without additional training. However, merged models are known to suffer from representation bias: systematic drift between the merged model's hidden states and those of each individual source model. Prior work (Yang et al., 2024a) study and mitigate this bias for encoder-based vision models using a lightweight correction module trained with L1 loss. However, such bias is not studied for decoder models due to their autoregressive nature. We analyze the problem of representation bias in decoder models, and show two challenges absent in encoders: (1) the causal attention mask causes bias to accumulate across token positions, requiring position-dependent correction; and (2) not all token positions are equally important, i.e., high-entropy (decision-critical) positions matter far more than low-entropy ones. To address these challenges, we propose Decoder-Aware Representation Tuning via Surgery (DARTS). DARTS employs a novel entropy-weighted L1 loss to upweight correction at high-entropy positions where errors most affect generation quality, and a per-position additive bias that captures position-dependent error without overparameterization. We perform extensive evaluation on three domains: code generation (HumanEval), mathematical reasoning (GSM8K), and instruction following (AlpacaEval) on Llama-2-7B models, and show DARTS achieves significant improvement over the standard surgery approach while adding negligible parameters ($0.1\%$ of total parameters).

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-31 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.28547 — 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 2026-10-17T10:50:25.819Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-31?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).