arXiv AI October 28, 2026 By Saket Reddy, Amro Abbas, Anshuman Suri, Ari Morcos, David Schwab, Pratyush Maini

Follow the Gradient: Tracing Safety Degradation Back to Training Data

The paper introduces SimpleLoRIF, a scalable gradient‑embedding technique that pinpoints fine‑tuning examples whose gradients align with unsafe behavior in large language models. By applying this method, the authors create SafeTulu, a version of Tulu 3 with the 20 % of most unsafe‑aligned examples removed, which reduces GCG attack success by up to 22 % across several LLMs while largely preserving or improving capability. The study also highlights previously under‑studied safety‑degrading data sources such as tool‑execution failures and narrative arithmetic.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

arXiv AI

Jul 3

Safety Targeted Embedding Exploit via Refinement

arXiv:2607. 01859v1 Announce Type: new Abstract: Safety training for large language models (LLMs) is conducted predominantly in English, leaving uncertain how well safety mechanisms generalize to low-resource languages and mixed-language code-switching.

By Joshua Adrian Cahyono

llms rag safety

Hugging Face Trending Papers

Jul 2

Safety Targeted Embedding Exploit via Refinement

Safety training for large language models (LLMs) is conducted predominantly in English, leaving uncertain how well safety mechanisms generalize to low-resource languages and mixed-language code-switching. We show that this creates an epistemic gap in which models confidently generate harmful responses for inputs that fall outside the distribution of their safety training.

llms rag safety

arXiv AI

Jun 2

DataShield: Safety-degrading Data Filtering for LLM Benign Instruction Fine-Tuning

arXiv:2606. 00160v1 Announce Type: cross Abstract: Large language models (LLMs) suffer from degraded safety capabilities even when fine-tuned with benign datasets.

By Junbo Zhang, Qianli Zhou, Xinyang Deng, Wen Jiang, Jie Pan, Jinbiao Zhu

llms nlp fine-tuning safety

arXiv Machine Learning

Jun 4

Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning

arXiv:2603. 07445v2 Announce Type: replace-cross Abstract: Large language models (LLMs) often require fine-tuning (FT) to perform well on downstream tasks, but FT can induce safety-alignment drift even when the training dataset contains only benign data.

By Guoli Wang, Haonan Shi, Tu Ouyang, An Wang

llms fine-tuning safety

arXiv AI

Sep 1

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

arXiv:2508.20766v2 Announce Type: replace-cross Abstract: Safety alignment in Large Language Models (LLMs) often involves mediating internal representations to refuse harmful requests. Recent researc...

By Harethah Abu Shairah, Hasan Abed Al Kader Hammoud, George Turkiyyah, Bernard Ghanem

llms fine-tuning benchmarks safety

arXiv Machine Learning

Oct 8

A Few Steps Further: Why Defenses Against Malicious Finetuning Erode Under Continued Training

arXiv:2605.14605v3 Announce Type: replace-cross Abstract: Model providers increasingly release the weights of large language models. Although these models are safety-aligned before release, their saf...

By Itay Zloczower, Eyal Lenga, Gilad Gressel, Yisroel Mirsky

llms fine-tuning safety

One email a morning, machine-written

Related stories

Safety Targeted Embedding Exploit via Refinement

Safety Targeted Embedding Exploit via Refinement

DataShield: Safety-degrading Data Filtering for LLM Benign Instruction Fine-Tuning

Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning

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

A Few Steps Further: Why Defenses Against Malicious Finetuning Erode Under Continued Training