mlsanityData to 14 Oct 2026

Why Do Some Language Models Fake Alignment While Others Don't?

arXiv 2506.18032 ↗First posted 22 Jun 2025cs.LGcited by 20

Not on the map. Why?

Abstract (arXiv)

Alignment faking in large language models presented a demonstration of Claude 3 Opus and Claude 3.5 Sonnet selectively complying with a helpful-only training objective to prevent modification of their behavior outside of training. We expand this analysis to 25 models and find that only 5 (Claude 3 Opus, Claude 3.5 Sonnet, Llama 3 405B, Grok 3, Gemini 2.0 Flash) comply with harmful queries more when they infer they are in training than when they infer they are in deployment. First, we study the motivations of these 5 models. Results from perturbing details of the scenario suggest that only Claude 3 Opus's compliance gap is primarily and consistently motivated by trying to keep its goals. Second, we investigate why many chat models don't fake alignment. Our results suggest this is not entirely due to a lack of capabilities: many base models fake alignment some of the time, and post-training eliminates alignment-faking for some models and amplifies it for others. We investigate 5 hypotheses for how post-training may suppress alignment faking and find that variations in refusal behavior may account for a significant portion of differences in alignment faking.

Key facts. Cited by 20 papers in the map (data through 14 Oct 2026). First posted on arXiv 22 Jun 2025.

Builds on

0 of 0 references.

No references here yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 4; 2026: 16.

20 papers.

Sort:

By area:

  1. 2026Constitutional Black-Box Monitoring for Scheming in LLM Agentscited by 10
  2. 2026Propensity Inference: Environmental Contributors to LLM Behaviourcited by 7
  3. 2025Toward Honest Language Models for Deductive Reasoningcited by 3
  4. 2026Value-Conflict Diagnostics Reveal Widespread Alignment Faking in Language Modelscited by 3
  5. 2026Towards Robust LLM Post-Training: Automatic Failure Management for Reinforcement Fine-Tuningcited by 3
  6. 2025Jinx: Unlimited LLMs for Probing Alignment Failurescited by 2
  7. 2026Decoding Hidden Deception in Reasoning LLMs: Activation Explainers for Deception Auditingcited by 2
  8. 2025Debate with Images: Detecting Deceptive Behaviors in Multimodal Large Language Modelscited by 1
  9. 2026Tatemae: Detecting Alignment Faking via Tool Selection in LLMscited by 1
  10. 2026Solipsistic Superintelligence is Unlikely to be Cooperativecited by 1
  11. 2026Model Forensics: Investigating Whether Concerning Behavior Reflects Misalignmentcited by 1
  12. 2025Corrigibility Transformation: Constructing Goals That Accept Updatescited by 0
8 more
  1. 2026Behavioural Analysis of Alignment Fakingcited by 0
  2. 2026Sycophancy Towards Researchers Drives Performative Misalignmentcited by 0
  3. 2026Order Is Not Control: Driven-Dissipative Response Laws Across Artificial and Biological Systemscited by 0
  4. 2026Defeat Devices in AI Systemscited by 0
  5. 2026Do Models Fake Alignment Without Clear Consequences?cited by 0
  6. 2026SchemeArena: Factorized Stress Testing of Scheming in LLM Agentscited by 0
  7. 2026Faking Good and Faking Bad in LLMs: Response Distortion Across Dark Triad Personality Traitscited by 0
  8. 2026Emergent Unfaithfulness: How Alignment Training Causes Language Models to Silently Override Task Faithfulnesscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Alignment faking in large language models
  2. 2025Empirical Evidence for Alignment Faking in a Small LLM and Prompt-Based Mitigation Techniques
  3. 2024Poser: Unmasking Alignment Faking LLMs by Manipulating Their Internals
  4. 2026The Alignment Paradox: How Post-Training Amplifies Confident Hallucinations in Language Models
  5. 2025Alignment Faking - the Train -> Deploy Asymmetry: Through a Game-Theoretic Lens with Bayesian-Stackelberg Equilibria
  6. 2023Fundamental Limitations of Alignment in Large Language Models
  7. 2025TRACEALIGN -- Tracing the Drift: Attributing Alignment Failures to Training-Time Belief Sources in LLMs
  8. 2025Eliciting and Analyzing Emergent Misalignment in State-of-the-Art Large Language Models