arXiv AI
Beyond Token-Level Guidance: Inference-Time Alignment of Specialized LLMs via Cross-Family Representation Steering
Read the original on arXiv AI →The Flow has not summarised this story yet — read it at arXiv AI.
Related stories
arXiv Machine Learning
Jun 4
Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning
arXiv:2603. 07445v2 Announce Type: replace-cross Abstract: Large language models (LLMs) often require fine-tuning (FT) to perform well on downstream tasks, but FT can induce safety-alignment drift even when the training dataset contains only benign data.
arXiv AI
Jun 11
ALIGNBEAM : Inference-Time Alignment Transfer via Cross-Vocabulary Logit Mixing
arXiv:2606. 12342v1 Announce Type: cross Abstract: Domain fine-tuning degrades the safety of large language models: fine-tuned specialists readily comply with harmful prompts framed in domain language.
Hugging Face Trending Papers
Jun 10
ALIGNBEAM : Inference-Time Alignment Transfer via Cross-Vocabulary Logit Mixing
Domain fine-tuning degrades the safety of large language models: fine-tuned specialists readily comply with harmful prompts framed in domain language. Existing inference-time defenses that mix logits from a safe anchor model require both models to share a vocabulary, which rules them out for the cross-family specialists where safety is most degraded.
Hugging Face Trending Papers
Sep 28
EOPSA: Efficient On-Policy Self-Distilled Safety Alignment
EOPSA (Efficient On-Policy Self-Distilled Safety Alignment) addresses inefficiencies in On-Policy Self-Distillation (OPSD) for safety alignment by focusing training on safety-critical tokens. It introduces Adaptive Rollout Scheduling, which limits generation length based on a Teacher Rescue Rate metric, and Selective Distillation, which filters out safety-neutral tokens to concentrate gradient updates on safety-pivotal transitions. Experiments on models up to 32B parameters show that EOPSA reduces rollout computation by about 50% and backpropagates through only roughly 2% of tokens, outperforming full-token distillation baselines in safety compliance and reasoning retention.
arXiv AI
Oct 31
Prompt-agnostic Language Models
The paper introduces Prompt-agnostic Language Models, focusing on train‑time techniques to embed prompt robustness during instruction tuning. It shows that mixing paraphrases in training causes significant gradient interference—57% of per‑paraphrasing gradients conflict—leading to the development of two gradient‑merging methods, Prompt‑TIES and Prompt‑DARE, which resolve this interference. Experiments across four models demonstrate that these methods substantially reduce per‑example variance and improve local robustness with minimal impact on overall accuracy.
arXiv AI
Jun 2
SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment
arXiv:2606. 02530v1 Announce Type: new Abstract: Aligning Large Language Models (LLMs) with human values often degrades their general capabilities, termed the alignment tax.