通过梯度引导优化序列生成模型的指令遵循能力。
Enhancing Instruction-Following Capabilities in Seq2Seq Models: DoLA Adaptations for T5
- 在T5中间解码层注入指令合规方向，提升模型响应一致性。
- 使记忆陷阱测试准确率从52%提升至99.7%。
- 适合研究大模型推理机制与可控生成的读者。
如FLAN-T5等编码器-解码器模型虽经指令微调，但当指令与训练中固化的内容冲突时仍易出错。我们为FLAN-T5引入DoLa方法，分析解码器表征演化过程。结果发现，T5中间层受编码器交叉注意力驱动发生快速变化，经语言建模头投影后各层呈现高度不稳定的词元偏好，导致对比解码不可靠。为此，我们提出基于梯度的激活引导方法，在解码中间层注入“指令合规”方向，此时表征既具语义意义又具可塑性。该干预显著提升记忆陷阱（MemoTrap）性能（52% → 99.7%），证明机制性引导在序列生成架构中可超越对比解码失效的问题。
原文摘要 · Abstract (English)
Encoder-decoder models such as FLAN-T5 are finetuned to follow instructions, but often fail when the instructions conflict with memorized continuations ingrained during training. To understand this behavior, we adapt DoLa to FLAN-T5 and examine how representations evolve in the decoder. Our findings show that T5's intermediate layers undergo rapid shifts driven by cross-attention to the encoder. When projected through the language modeling head, each depth presents highly volatile token preferences, leading to unreliable behavior with contrastive decoding. Motivated by this, we introduce a gradient-based activation-steering method that injects an "instruction-compliance" direction into mid-decoder layers, where the representation is both meaningful and still malleable. This intervention dramatically improves MemoTrap performance (52% to 99.7%), demonstrating that mechanistic steering can succeed where contrastive decoding fails in Seq2Seq architectures.
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。