Pith. sign in

REVIEW 8 cited by

Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2402.11746 v1 pith:7G2Q365I submitted 2024-02-19 cs.CL cs.AI

classification cs.CLcs.AI
keywords restasafetyarithmeticcompromisedfine-tuninglanguagemodelmodels
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Aligned language models face a significant limitation as their fine-tuning often results in compromised safety. To tackle this, we propose a simple method RESTA that performs LLM safety realignment. RESTA stands for REstoring Safety through Task Arithmetic. At its core, it involves a simple arithmetic addition of a safety vector to the weights of the compromised model. We demonstrate the effectiveness of RESTA in both parameter-efficient and full fine-tuning, covering a wide range of downstream tasks, including instruction following in Chinese, English, and Hindi, as well as problem-solving capabilities in Code and Math. We also showcase the generalizability of RESTA on three existing safety evaluation benchmarks and a multilingual benchmark dataset proposed as a part of this work, consisting of 550 harmful questions covering 11 categories, each with 5 sub-categories of harm. Overall, RESTA decreases the harmfulness of the compromised model from 18.6% to 5.1% and from 9.2% to 1.5% in parameter-efficient and full fine-tuning, respectively, while maintaining most of the model's performance on the task. We release the source codes at: https://github.com/declare-lab/resta.

Discussion (0). Sign in to comment.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Understanding Refusal in Language Models with Sparse Autoencoders

    cs.CL 2025-05 conditional novelty 7.0 of 10

    Refusal in Gemma-2-2B and Llama-3.1-8B is mediated by a small set of SAE features, harm features causally activate refusal features, and adversarial jailbreaks suppress those refusal features.

  2. MoGU V2: Toward a Higher Pareto Frontier Between Model Usability and Security

    cs.CL 2025-09 conditional novelty 5.0 of 10

    MoGUv2 embeds small routers in the deeper layers of LLMs to dynamically blend a helpful variant and a refusal variant, improving safety against jailbreak and fine-tuning attacks while preserving usability.

  3. Anchoring Refusal Direction: Mitigating Safety Risks in Tuning via Projection Constraint

    cs.CL 2025-09 conditional novelty 5.0 of 10

    ProCon anchors each sample's hidden-state projection onto the LLM's initial refusal direction during instruction fine-tuning, reducing refusal-direction drift and safety risks with limited task-performance loss.

  4. Turning the Spell Around: Lightweight Alignment Amplification via Rank-One Safety Injection

    cs.CL 2025-08 conditional novelty 5.0 of 10

    ROSI bakes the refusal direction into a model's weight matrices via a rank-one update, raising refusal and jailbreak robustness with minimal measured utility cost.

  5. SEALGuard: Safeguarding the Multilingual Conversations in Southeast Asian Languages for LLM Software Systems

    cs.CL 2025-07 conditional novelty 5.0 of 10

    A LoRA-adapted SeaLLM model detects unsafe and jailbreak prompts in nine Southeast Asian languages with 97% recall and 98% F1 on the authors' new SEALSBench benchmark, far above zero-shot LlamaGuard.

  6. Three Minds, One Legend: Jailbreak Large Reasoning Model with Adaptive Stacked Ciphers

    cs.CL 2025-05 conditional novelty 5.0 of 10

    A stacked-cipher jailbreak with adaptive code selection achieves 80.8% to 100% attack success on commercial large reasoning models.

  7. SafeSteer: Interpretable Safety Steering with Refusal-Evasion in LLMs

    cs.LG 2025-06 conditional novelty 4.0 of 10

    SafeSteer uses category-specific activation vectors to steer LLMs toward safe, on-topic, non-refusing responses at inference time.

  8. Unlocking the Effectiveness of LoRA-FP for Seamless Transfer Implantation of Fingerprints in Downstream Models

    cs.CR 2025-08 conditional novelty 3.0 of 10

    Backdoor fingerprints trained into LoRA adapters on a base LLM transfer to derivative models with 100% trigger success and, in several scenarios, greater robustness than directly injected fingerprints.

Pith tools