Papers
Topics
Authors
Recent
AI Research Assistant
AI Research Assistant
Well-researched responses based on relevant abstracts and paper content.
Custom Instructions Pro
Preferences or requirements that you'd like Emergent Mind to consider when generating responses.
Gemini 2.5 Flash
Gemini 2.5 Flash 85 tok/s
Gemini 2.5 Pro 46 tok/s Pro
GPT-5 Medium 16 tok/s Pro
GPT-5 High 10 tok/s Pro
GPT-4o 108 tok/s Pro
Kimi K2 192 tok/s Pro
GPT OSS 120B 455 tok/s Pro
Claude Sonnet 4 31 tok/s Pro
2000 character limit reached

Manipulating Transformer-Based Models: Controllability, Steerability, and Robust Interventions (2509.04549v1)

Published 4 Sep 2025 in cs.CL and cs.AI

Abstract: Transformer-based LLMs excel in NLP tasks, but fine-grained control remains challenging. This paper explores methods for manipulating transformer models through principled interventions at three levels: prompts, activations, and weights. We formalize controllable text generation as an optimization problem addressable via prompt engineering, parameter-efficient fine-tuning, model editing, and reinforcement learning. We introduce a unified framework encompassing prompt-level steering, activation interventions, and weight-space edits. We analyze robustness and safety implications, including adversarial attacks and alignment mitigations. Theoretically, we show minimal weight updates can achieve targeted behavior changes with limited side-effects. Empirically, we demonstrate >90% success in sentiment control and factual edits while preserving base performance, though generalization-specificity trade-offs exist. We discuss ethical dual-use risks and the need for rigorous evaluation. This work lays groundwork for designing controllable and robust LLMs.

Summary

We haven't generated a summary for this paper yet.

Lightbulb On Streamline Icon: https://streamlinehq.com

Continue Learning

We haven't generated follow-up questions for this paper yet.

List To Do Tasks Checklist Streamline Icon: https://streamlinehq.com

Collections

Sign up for free to add this paper to one or more collections.

Reddit Logo Streamline Icon: https://streamlinehq.com