mlsanityData to 7 Oct 2026

Safeguard-Conditioned Uplift: Measuring Utility-Risk Frontiers for Dual-Use Biology Assistants

arXiv 2607.13039 ↗First posted 12 Jun 2026cs.CY cs.AIcited by 1

Not on the map. Why?

Abstract (arXiv)

A refusal rate neither identifies which component intervened nor measures its burden on legitimate users. This paper evaluates safeguards for dual-use biology assistants at the action and answer levels. The framework reconstructs the access path, separates provider refusals from downstream actions, and selects thresholds under an intervention budget. A frozen fresh-generation study satisfies its criterion on Claude Opus 4.5, but both passing configurations share one upstream provider effect; none passes on Gemini 2.5 Flash. The fixed Opus policies retain positive selectivity on 104 previously unused released-label pairs, but both fail a 20% matched-benign constraint. At the answer level, no joint-scoring verifier qualifies on a response-disjoint 7,200-judgment holdout. A fresh 8,640-judgment factorial experiment finds separate gains from criterion isolation and ordinal representation, with a positive interaction between them; requiring explicit localization lowers aggregate accuracy under a strict no-repair schema. The evidence supports prospective action-level selectivity and identifies verifier interface effects, but not calibrated selective access, verified content removal, or biological-risk reduction.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 34 of its 47 references that are in the map. First posted on arXiv 12 Jun 2026.

Builds on

34 of 47 references.

References · 34

  1. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  2. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 2,914
  3. 2023G-Eval: NLG Evaluation using GPT-4 with Better Human Alignmentcited by 2,205
  4. 2023Jailbroken: How Does LLM Safety Training Fail?cited by 1,565
  5. 2024HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusalcited by 1,330
  6. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,235
  7. 2022Red Teaming Language Models with Language Modelscited by 983
  8. 2023BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Datasetcited by 828
  9. 2023AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Modelscited by 738
  10. 2024JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Modelscited by 494
  11. 2024The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearningcited by 488
  12. 2023XSTest: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Modelscited by 477
22 more
  1. 2024WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMscited by 460
  2. 2022ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detectioncited by 419
  3. 2024A StrongREJECT for Empty Jailbreakscited by 352
  4. 2021Distribution-Free, Risk-Controlling Prediction Setscited by 278
  5. 2021Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Controlcited by 256
  6. 2024Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judgecited by 249
  7. 2024SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Modelscited by 225
  8. 2023ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversationcited by 224
  9. 2024OR-Bench: An Over-Refusal Benchmark for Large Language Modelscited by 222
  10. 2023SafetyBench: Evaluating the Safety of Large Language Modelscited by 189
  11. 2024Rainbow Teaming: Open-Ended Generation of Diverse Adversarial Promptscited by 189
  12. 2025Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red Teamingcited by 172
  13. 2024LAB-Bench: Measuring Capabilities of Language Models for Biology Researchcited by 138
  14. 2024SG-Bench: Evaluating LLM Safety Generalization Across Diverse Tasks and Prompt Typescited by 58
  15. 2024On Evaluating the Durability of Safeguards for Open-Weight LLMscited by 51
  16. 2026Constitutional Classifiers++: Efficient Production-Grade Defenses against Universal Jailbreakscited by 36
  17. 2025Virology Capabilities Test (VCT): A Multimodal Virology Q&A Benchmarkcited by 27
  18. 2025Multi-lingual Multi-turn Automated Red Teaming for LLMscited by 12
  19. 2026Measuring Mid-2025 LLM-Assistance on Novice Performance in Biologycited by 5
  20. 2026ABC-Bench: An Agentic Bio-Capabilities Benchmark for Biosecuritycited by 5
  21. 2026LLM Novice Uplift on Dual-Use, In Silico Biology Taskscited by 4
  22. 2026Evaluating Human-AI Safety: A Framework for Measuring Harmful Capability Upliftcited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026The Safeguard Worked. Is the LLM System Safer?cited by 0