TALAN让大模型在微调时精准增强推理能力，不伤原有优势。
TALAN: Task-Aligned Latent Adaptation Networks for Targeted Post-Training of Large Language Models
- 在残差流中插入可调潜空间路径，联合训练低秩适配器
- 相比基线平均提升1.41分（LoRA）和1.85分（DoRA），全模型均正向
- 参数开销小于1%，适合高效部署与激活级调控研究
目标后训练旨在提升大模型的推理、数学与代码能力，同时避免削弱原有优势。低秩适配器效率高但全局适用；激活干预具输入感知性，却常需额外探针或推理时控制。本文提出TALAN（任务对齐潜空间适配网络），一种序列条件化的潜空间侧路径，插入Transformer残差流并同低秩适配器在一次SFT循环中联合训练。TALAN将活跃序列压缩为潜记忆，重混为逐标记扰动，并通过可控残差更新写回。其配置涵盖六个维度：插入位置、记忆大小、混合器、写回规则、可训练范围与梯度尺度。在四个Qwen3家族主干模型及四组STEM/代码基准上，TALAN优于匹配的LoRA与DoRA基线。采用LoRA时，跨模型平均增益+1.41点，所有四主干均正向，16个模型-基准组合中无负值。采用DoRA时，平均增益+1.85点，所有主干为正，16个中13个为正。成对种子测试支持平均正向效应，但存在显著方差，视为敏感性分析。成本极低：可训练参数占比<1%，推理开销仅1.01-1.02倍于匹配的LoRA。Llama-3.2-1B迁移探针在七组种子下也呈正向，支持跨模型迁移。内部状态分析表明，TALAN是小规模互补激活干预：适配器更新量为TALAN扰动的80–1,700倍，但方向近似正交；分层测量显示该微小正交扰动可沿深度传播并放大。TALAN为标准适配器后训练中的可调控激活级适配提供实用平台。
原文摘要 · Abstract (English)
Targeted post-training aims to improve reasoning, math, and code without degrading strengths. Low-rank adapters are efficient but task-global; activation interventions are input-aware but often require separate probes, vectors, or inference-time steering. We introduce TALAN (Task-Aligned Latent Adaptation Networks), a sequence-conditioned latent side path inserted into a transformer's residual stream and co-trained with a low-rank adapter in one SFT loop. TALAN compresses the active sequence into latent memory, remixes it into token-level perturbations, and writes them back through a controlled residual update. It is configured along six axes: insertion location, memory size, mixer, writeback rule, trainability scope, and gradient scale. Across four Qwen3-family backbones and four STEM/code benchmarks, TALAN improves matched LoRA and DoRA baselines. With LoRA, it yields a +1.41 point cross-model mean gain, positive on all four backbones and non-negative on all 16 model-benchmark cells. With DoRA, it yields a +1.85 point mean gain, positive on all backbones and on 13 of 16 cells. Paired seed checks support positive average effects but show nontrivial variance, so we treat them as sensitivity checks. Cost is small: <1% trainable parameters relative to the backbone and 1.01-1.02x inference overhead versus matched LoRA. A Llama-3.2-1B transfer probe is also positive under LoRA and rsLoRA across seven paired seeds, supporting a transfer beyond Qwen. Internal-state analyses suggest TALAN is a small complementary activation intervention. The matched adapter update is 80-1,700x larger than the TALAN perturbation, yet their directions have near-zero cosine; per-layer measurements show this small orthogonal perturbation propagates and amplifies through depth. TALAN offers a practical platform for studying steerable activation-level adaptation within standard adapter-based post-training.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。