关注这个，而非那个！通过自适应特征指定引导大语言模型
Focus On This, Not That! Steering LLMs with Adaptive Feature Specification
AI总结：
本文提出 Focus Instruction Tuning（FIT），通过训练大语言模型在响应时关注指定特征并忽略其他特征，实现推理时行为引导、鲁棒性提升、社会偏见缓解及分布偏移下的泛化，为构建更可控的 LLM 提供轻量级内在机制。
摘要：
尽管指令微调（IT）在训练大语言模型（LLMs）方面取得了成功，但此类模型往往会利用从训练数据中学到的虚假或偏置特征，并可能变得失准，从而导致不期望的行为。虽然现有技术可以在推理时引导模型行为，但它们通常是事后补救性的，并未将引导作为模型的内在特征嵌入。在这项工作中，我们提出了 Focus Instruction Tuning（FIT），它训练大语言模型通过关注特定特征同时忽略其他特征来条件化其响应，从而根据指定的特征产生不同的行为。在多个不同的基准测试上，我们证明了 FIT：（i）能在推理时成功引导行为；（ii）通过放大核心任务信号并降低虚假线索的权重来提高鲁棒性；（iii）通过抑制人口统计属性来减轻社会偏见；（iv）在分布偏移下以及对先前未见过的关注特征具有泛化能力。因此，FIT 为构建更鲁棒、更公平且易于控制的大语言模型提供了一种轻量级的内在机制。
英文摘要：
Despite the success of Instruction Tuning (IT) in training large language models (LLMs), such models often leverage spurious or biased features learnt from their training data and can become misaligned, leading to undesired behaviours. While existing techniques can steer model behaviour at inference-time, they are often post-hoc and do not embed steering as an intrinsic model feature. In this work, we introduce Focus Instruction Tuning (FIT), which trains LLMs to condition their responses by focusing on specific features whilst ignoring others, leading to different behaviours based on what features are specified. Across diverse benchmarks, we demonstrate that FIT: (i) successfully steers behaviour at inference time; (ii) increases robustness by amplifying core task signals and down-weighting spurious cues; (iii) mitigates social bias by suppressing demographic attributes; and (iv) generalises under distribution shifts and to previously unseen focus features. FIT therefore offers a lightweight, intrinsic mechanism for building more robust, fair, and easily controllable LLMs.
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。
