从频谱显著性到有限删除:评估选择性语言模型编辑
From Spectral Saliency to Finite Deletion: Evaluating Selective Language Model Edits
AI总结:
本研究针对选择性语言模型编辑,提出将误差分解为导数传递与非线性响应的方法,通过区间筛选评估删除效应,在Mistral等模型上验证了区间覆盖等属性需单独评估。
摘要:
删除奇异组件是从语言模型中移除某项能力的精确方式,然而用于对组件排序的一阶频谱显著性仅能预测局部损失变化,而删除是一种有限干预。本文研究预测误差如何影响选择性编辑,其中选择性编辑要求在降低目标任务性能的同时保留其他任务性能。我们将误差分解为跨项拆分的导数传递和沿删除路径的非线性响应。我们证明,当每个候选的区间包含其在所有任务上的有限验证损失变化时,外部区间筛选会保留所有可行候选。我们使用仅含答案的负对数似然目标来评估这些效应,该目标用于搜索、验证和测试,且采用不重叠的校准和评估掩码。主要配置针对Mistral-7B-Instruct-v0.3模型的算术能力进行删除,同时保留代码和问答能力;扩展研究覆盖Llama-3.1-8B-Instruct模型及反向目标角色。在主要研究中,从7168个支撑中删除1024个坐标,校准区间覆盖1024个评估掩码中的968个。外部筛选在25次验证检查后得到直接选定的编辑,而直接验证在32次检查上限时弃权(不执行),但拒绝了74个可行掩码中的4个。所有三个选定的编辑均通过验证,但仅一个满足所有测试约束。因此,区间覆盖、可行候选保留和独立测试保留是不同的属性,需要单独评估。
英文摘要:
Deleting singular components offers a precise way to remove a capability from a language model, yet the first-order spectral saliency used to rank components predicts only a local loss change, whereas deletion is a finite intervention. In this paper, we study how prediction error affects selective editing, where selective editing requires degradation of a target task while preserving others. We decompose the error into derivative transfer across item splits and nonlinear response along the deletion path. We establish that outer interval screening retains every feasible candidate when every candidate's intervals contain its finite validation loss changes on every task. We evaluate these effects using one answer-only negative log-likelihood objective for search, validation, and test, with disjoint calibration and evaluation masks. The primary configuration targets arithmetic in Mistral-7B-Instruct-v0.3 while preserving code and question answering; extensions cover Llama-3.1-8B-Instruct and reversed target roles. In the primary study, which deletes 1024 coordinates from a support of 7168, calibrated intervals cover 968 of 1024 evaluation masks. Outer screening reaches the direct-selected edit after 25 verification checks, whereas direct verification abstains at a 32-check cap, but it rejects 4 of 74 feasible masks. All three selected edits pass validation, yet only one meets every test constraint. Interval coverage, feasible-candidate retention, and independent test preservation are therefore distinct properties and require separate assessment.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。