mlsanityData to 14 Oct 2026

Quant Fever, Reasoning Blackholes, Schrodinger's Compliance, and More: Probing GPT-OSS-20B

arXiv 2509.23882 ↗First posted 28 Sep 2025cs.AI cs.CRcited by 1

On the map: LLM reasoning → Reasoning for agents, planning and code

Abstract (arXiv)

OpenAI's GPT-OSS family provides open-weight language models with explicit chain-of-thought (CoT) reasoning and a Harmony prompt format. We summarize an extensive security evaluation of GPT-OSS-20B that probes the model's behavior under different adversarial conditions. Using the Jailbreak Oracle (JO) [1], a systematic LLM evaluation tool, the study uncovers several failure modes including quant fever, reasoning blackholes, Schrodinger's compliance, reasoning procedure mirage, and chain-oriented prompting. Experiments demonstrate how these behaviors can be exploited on the GPT-OSS-20B model, leading to severe consequences.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning. Builds on 6 of its 12 references that are in the map. First posted on arXiv 28 Sep 2025.

Builds on

6 of 12 references.

References · 6

  1. 2025gpt-oss-120b & gpt-oss-20b Model Cardcited by 1,386
  2. 2024JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Modelscited by 536
  3. 2024AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agentscited by 437
  4. 2024A StrongREJECT for Empty Jailbreakscited by 385
  5. 2022Why Should Adversarial Perturbations be Imperceptible? Rethink the Research Paradigm in Adversarial NLPcited by 102
  6. 2025Toward Principled LLM Safety Testing: Solving the Jailbreak Oracle Problemcited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Control-Token Injection Suppresses Chain-of-Thought and Defeats Reasoning-Based Oversight in Tool-Using Agentscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Process or Result? Manipulated Ending Tokens Can Mislead Reasoning LLMs to Ignore the Correct Reasoning Steps
  2. 2025H-CoT: Hijacking the Chain-of-Thought Safety Reasoning Mechanism to Jailbreak Large Reasoning Models, Including OpenAI o1/o3, DeepSeek-R1, and Gemini 2.0 Flash Thinking
  3. 2025An Empirical Study on the Security Vulnerabilities of GPTs ↗
  4. 2025All Code, No Thought: Current Language Models Struggle to Reason in Ciphered Language
  5. 2026Fragile Thoughts: How Large Language Models Handle Chain-of-Thought Perturbations
  6. 2025AdvChain: Adversarial Chain-of-Thought Tuning for Robust Safety Alignment of Large Reasoning Models
  7. 2026Exposing LLM Safety Gaps Through Mathematical Encoding:New Attacks and Systematic Analysis
  8. 2025CipherBank: Exploring the Boundary of LLM Reasoning Capabilities through Cryptography Challenges