analiza aiqbench
This research introduces a modular framework using Activated LoRA adapters and context-aware routing to dynamically mitigate specific harms like bias and toxicity in Large Language Models during generation. By enabling mid-sequence corrections without invalidating the KV cache, the approach offers a scalable path to safer LLM deployments while preserving task performance.
An Efficient and Modular Framework for Targeted Harm Mitigation in LLMS
arXiv:2609.13624v1 Announce Type: cross Abstract: Large Language Models (LLMs) are powerful zero-shot learners but remain prone to misalignment with human preferences, often producing biased, toxic, or otherwise harmful outputs. Existing alignment methods, while effective, are costly and tightly coupled to the model, limiting flexibility and scalability. We propose a modular correction framework that augments pretrained LLMs with Activated LoRA (aLoRA) adapters and a context-aware routing mechanism to eliminate harms from misaligned model responses. Our approach enables expert adapters to activate mid-sequence without invalidating the KV cache, allowing low-latency, targeted correction during generation. Each expert is trained to detect and mitigate specific harms, such as bias or toxicity. A learned router dynamically selects appropriate experts based on the models intermediate outputs. We demonstrate that our system improves alignment on standard safety benchmarks while preserving task performance, offering a lightweight and efficient path toward safer and more controllable LLM deployments.
Abstract:Large Language Models (LLMs) are powerful zero-shot learners but remain prone to misalignment with human preferences, often producing biased, toxic, or otherwise harmful outputs. Existing alignment methods, while effective, are costly and tightly coupled to the model, limiting flexibility and scalability. We propose a modular correction framework that augments pretrained LLMs with Activated LoRA (aLoRA) adapters and a context-aware routing mechanism to eliminate harms from misaligned model responses. Our approach enables expert adapters to activate mid-sequence without invalidating the KV cache, allowing low-latency, targeted correction during generation. Each expert is trained to detect and mitigate specific harms, such as bias or toxicity. A learned router dynamically selects appropriate experts based on the models intermediate outputs. We demonstrate that our system improves alignment on standard safety benchmarks while preserving task performance, offering a lightweight and efficient path toward safer and more controllable LLM deployments.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY)
arXiv:2609.13624 [cs.LG]
(or arXiv:2609.13624v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.13624 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 248 cuvinte
Subiecte conexe
The J-Space Debate, Agent Swarms, and Pacing Frontier AI - Digital Minds Newsletter #4
LessWrong · 18 sept. 2026
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · 5 oct. 2026
Does AI Reviewer See the Full Picture? Attacking and Defending Multimodal Peer Review
arXiv · 12 iun. 2026
[AINews] not much happened today
Latent Space · 18 sept. 2026