On the Effectiveness-Fluency Trade-Off in LLM Conditioning: A Systematic Study A systematic study by Iuri Macocco, Pau Rodríguez Lopez, Arno Blaas, Luca Zappella, Marco Baroni and Xavier Suau Cuadros, published in Transactions on Machine Learning Research (TMLR) in September 2026, found that efficient activation steering methods achieve concept conditioning at a steep cost to fluency and are far less effective on instruction-tuned models than on their base counterparts. The authors report that simple prompting and full supervised fine-tuning are viable for concept injection but weaker at concept removal, and that cheaply computed textual metrics highly correlate with costly LLM-as-judge scores. content type paper https://machinelearning.apple.com/research/ published September 2026 On the Effectiveness-Fluency Trade-Off in LLM Conditioning: A Systematic Study AuthorsIuri Macocco†, Pau Rodríguez Lopez, Arno Blaas, Luca Zappella, Marco Baroni† , Xavier Suau Cuadros Controlling the output of Large Language Models LLMs is a central challenge for their reliable deployment, yet a clear understanding of the involved trade-offs remains elusive. Current approaches to conditioning are often evaluated with a narrow focus on their effectiveness at injecting or removing a target concept, neglecting generation quality. We systematically investigate a range of conditioning methods in both injection and removal scenarios. We find that efficient steering methods frequently achieve conditioning at a steep cost to fluency. Furthermore, we identify a critical yet previously overlooked interaction with the training paradigm: activation steering methods are far less effective on instruction-tuned models than on their base counterparts. Simple prompting and full-fledged supervised fine-tuning, on the other hand, are viable options for concept injection, but are not as good at concept removal. Finally, cheaply computed textual metrics highly correlate to costly LLM-as-judge scores, and provide insights on the behavior of conditioning methods. Dynamically Scaled Activation Steering September 18, 2026 research area Human-Computer Interaction https://machinelearning.apple.com/research/?domain=Human-Computer%20Interaction , research area Methods and Algorithms https://machinelearning.apple.com/research/?domain=Methods%20and%20Algorithms Transactions on Machine Learning Research TMLR