paper
On The Effectiveness-Fluency Trade-Off In LLM Conditioning: A Systematic Study
paperactiveprovisional
on-the-effectiveness-fluency-trade-off-in-llm-conditioning-a-systematic-study-dcfa5cf6·1 events·first seen 6d agoAliases: On The Effectiveness-Fluency Trade-Off In LLM Conditioning: A Systematic Study
More like this (12)
Leveraging Audio-LLMs to Filter Speech-to-Speech Training DataWhen English Isn't the Best Teacher: Source Language Effects in Cross-Lingual In-Context LearningJanus: A Benchmark for Goal-Conditioned Information Distortion in LLMsEDIT: Evidence-Diagnosed Intervention Training for Rule-Faithful LLM GradingPC Layer: Polynomial Weight Preconditioning for Improving LLM Pre-TrainingBe My Tutor: On-Policy Co-Distillation for Mutual LLM Improvement via Peer FeedbackA sleep-like consolidation mechanism for LLMsExpRL: Exploratory RL for LLM Mid-TrainingAttention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix ItLearning to Hear Hesitation: Continual Learning for Disfluency-Aware ASRContinual LLM Upcycling: A Predictor-Gated Bank-Wise Sparsity Training Recipe for Dense-to-Sparse LLMsTraining LLMs to Enforce Multi-Level Instruction Hierarchies via Gravity-Weighted Direct Preference Optimization
Recent events (1)
Systematic study reveals effectiveness-fluency trade-offs in LLM conditioning methods
A new arXiv paper systematically evaluates a range of LLM conditioning methods across both concept injection and removal scenarios, finding that efficient steering methods often degrade fluency significantly. A key finding is that activation steering is substantially less effective on instruction-tuned models than on base models, a previously overlooked interaction. Simple prompting and supervised fine-tuning work for concept injection but not removal, and cheap textual metrics are found to correlate well with expensive LLM-as-judge evaluations.