Pith. sign in

REVIEW 4 cited by

Model Merging and Safety Alignment: One Bad Model Spoils the Bunch

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 2406.14563 v1 pith:ERDOZTRZ submitted 2024-06-20 cs.CL cs.AIcs.LG

classification cs.CLcs.AIcs.LG
keywords mergingmodelalignmentdataexpertisemodelssafetydomain
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

Merging Large Language Models (LLMs) is a cost-effective technique for combining multiple expert LLMs into a single versatile model, retaining the expertise of the original ones. However, current approaches often overlook the importance of safety alignment during merging, leading to highly misaligned models. This work investigates the effects of model merging on alignment. We evaluate several popular model merging techniques, demonstrating that existing methods do not only transfer domain expertise but also propagate misalignment. We propose a simple two-step approach to address this problem: (i) generating synthetic safety and domain-specific data, and (ii) incorporating these generated data into the optimization process of existing data-aware model merging techniques. This allows us to treat alignment as a skill that can be maximized in the resulting merged LLM. Our experiments illustrate the effectiveness of integrating alignment-related data during merging, resulting in models that excel in both domain expertise and alignment.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

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

  1. Consiglieres in the Shadow: Understanding the Use of Uncensored Large Language Models in Cybercrimes

    cs.CR 2025-08 conditional novelty 7.0 of 10

    Using a knowledge graph of model and dataset relationships, this paper identifies more than 11,000 uncensored LLMs on Hugging Face and documents their use in cybercrime services.

  2. When Skills Meet Safety: Benchmarking and Characterizing the Adaptive Jailbreak Robustness of Skill-Merged LLMs

    cs.LG 2026-08 accept novelty 6.0 of 10

    Static refusal tests overstate the safety of skill-merged LLMs: models with identical static safety differ sharply under adaptive attack, and a task-vector overlap with a safety subspace flags only same-recipe abliter...

  3. On Almost Surely Safe Alignment of Large Language Models at Inference-Time

    cs.LG 2025-02 conditional novelty 6.0 of 10

    An inference-time beam-search method with a safety-state tracker and latent critic enforces a user-supplied safety cost model, with an almost-sure guarantee only relative to that model.

  4. If You Can't Use Them, Recycle Them: Optimizing Merging at Scale Mitigates Performance Tradeoffs

    cs.CL 2024-12 conditional novelty 5.0 of 10

    Optimizing linear merge weights on 16 104B generalist checkpoints via CMA-ES reduces task tradeoffs and beats individual models and simple merge baselines.

Pith tools