arXiv cs.CL· Noor Islam S. Mohammad, Ulug Bayazit·· 14 小时前AI 评分8
SURGELLM：用任务感知特征门控与类别平衡归一化重新审视多任务评估
SURGELLM: Rethinking Multi-Task Evaluation through Task-Aware Feature Gating with Class-Balanced Normalization
AI 导读
SURGELLM 是一个统一 Transformer 框架，通过手术式特征门控、任务条件前缀 token 和实例加权归一化（IWN）三个轻量模块，解决异构 NLP 任务中的归纳偏置不匹配与类别不平衡问题。
正文 · 原文
This paper has been withdrawn by arXiv Admin
No PDF available, click to view other formats
Abstract:Fine-tuned encoders deployed across heterogeneous NLP tasks face three compounding problems: mismatched inductive biases, class-imbalance corruption of feature statistics, and no mechanism to condition attention on external lexical knowledge. We introduce \textbf{\surgellm}, a unified transformer framework that addresses each with a dedicated lightweight module: a \emph{surgical feature gate} (learned per-dimension sigmoid over curated lexical indicators and \texttt{[CLS]}; provably degenerates to identity when features are uninformative), \emph{task-conditioned prefix tokens} (quantized feature values and task identity prepended to every input), and \emph{Instance-Weighted Normalization} (IWN; removes class-prior bias from gate statistics). We prove an excess-risk bound linking gate benefit to \emph{surgical feature alignment}. Across four tasks, SST-2, multi-hop retrieval, LLM-prompt attribution, and authorship detection, covering 17,830 examples and eleven model variants over three seeds, the IWN variant achieves macro-F1 \textbf{0.940} ($+0.036$ over the strongest non-IWN baseline; $+0.130$ on authorship detection). A random-vocabulary control ($-0.028$ avg.\ F1) confirms gains are lexical, not parametric. Code, vocabularies, and a $99.5\%$-recovery auto-extraction recipe are released.
| Comments: | arXiv admin note: This submission has been withdrawn because it does not meet arXiv's research content quality standards |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.24259 [cs.CL] |
| (or arXiv:2606.24259v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.24259 arXiv-issued DOI via DataCite |
Submission history
From: arXiv Admin [view email]
[v1]
Tue, 23 Jun 2026 07:47:21 UTC (161 KB)
[v2]
Mon, 26 Oct 2026 15:31:47 UTC (1 KB) (withdrawn)
来源：arXiv cs.CL · arxiv.org