An Efficient and Modular Framework for Targeted Harm Mitigation in LLMS

Published 16 Sept 2026arXiv:2609.13624

data quality89

Updated 12 h ago · first seen 15 Sept 2026

paper_01M2JK0BQS3M8KENTASR0VS44T

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.

Authors

Authors 8

Kristjan GreenewaldMichal UlewiczMomin AbbassMuneeza AzmatNathalie BaracaldoRaya HoreshRoberto CampbellRog\'erio Abreu de Paula

Linked names open researcher pages (created from the paper's author list; name-only, no affiliation unless a source states it). Unlinked names have no researcher record yet.

Organizations

Organizations 0

No organization stated. arXiv metadata does not carry affiliations; an organization is linked only when a model card or lab page cites the paper.

Models

Models introduced or described 0

Inbound described_by relations from model cards and documentation.

No model links this paper yet

Model pages link papers through their model cards and documentation; the relation is written only when a source states it.

Datasets

Datasets used 0

No dataset relation recorded.

Benchmarks

Benchmarks used 0

No benchmark relation recorded.

Code

Repositories & frameworks 0

No repository linked.

Timeline

Timeline 3

16 Sept 2026
Property changedPaperAn Efficient and Modular Framework for Targeted Harm Mitigation in LLMS
An Efficient and Modular Framework for Targeted Harm Mitigation in LLMS: published at changed from 2026-09-15T04:00:00+00:00 to 2026-09-16T04:00:00+00:00
Published15 Sept 2026→16 Sept 2026arxiv
rss.arxiv.org
15 Sept 2026
Property changedPaperAn Efficient and Modular Framework for Targeted Harm Mitigation in LLMS
An Efficient and Modular Framework for Targeted Harm Mitigation in LLMS: arxiv announce type changed from new to cross
Arxiv announce typenew→crossarxiv
rss.arxiv.org
15 Sept 2026
New paperPaperAn Efficient and Modular Framework for Targeted Harm Mitigation in LLMS
New paper: An Efficient and Modular Framework for Targeted Harm Mitigation in LLMS
arxiv
rss.arxiv.org

Full timeline →

Sources

Sources 3

Source documents
Source	Document	Type	Tier	Last observed	Snapshots
arXiv (Atom API + RSS)	rss.arxiv.org/rss/cs.AI	feed	T1· Official	12 h ago	5
arXiv (Atom API + RSS)	rss.arxiv.org/rss/cs.CL	feed	T1· Official	12 h ago	4
arXiv (Atom API + RSS)	rss.arxiv.org/rss/cs.LG	feed	T1· Official	12 h ago	4

Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.