mlsanityData to 14 Oct 2026

AI Testing Should Account for Sophisticated Strategic Behaviour

arXiv 2508.14927 ↗First posted 19 Aug 2025cs.GT cs.AIcited by 6

Not on the map. Why?

Abstract (arXiv)

This position paper argues for two claims regarding AI testing and evaluation. First, to remain informative about deployment behaviour, evaluations need account for the possibility that AI systems understand their circumstances and reason strategically. Second, game-theoretic analysis can inform evaluation design by formalising and scrutinising the reasoning in evaluation-based safety cases. Drawing on examples from existing AI systems, a review of relevant research, and formal strategic analysis of a stylised evaluation scenario, we present evidence for these claims and motivate several research directions.

Key facts. Cited by 6 papers in the map (data through 14 Oct 2026). Builds on 34 of its 81 references that are in the map. First posted on arXiv 19 Aug 2025.

Builds on

34 of 81 references.

References · 34

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2017Deep reinforcement learning from human preferencescited by 3,912
  3. 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,411
  4. 2019Fine-Tuning Language Models from Human Preferencescited by 2,150
  5. 2019SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systemscited by 1,729
  6. 2019Mastering Atari, Go, Chess and Shogi by Planning with a Learned Modelcited by 1,353
  7. 2024Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Trainingcited by 529
  8. 2025Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscationcited by 384
  9. 2024Alignment faking in large language modelscited by 360
  10. 2023Poisoning Web-Scale Training Datasets is Practicalcited by 290
  11. 2024Frontier Models are Capable of In-context Schemingcited by 287
  12. 2019Risks from Learned Optimization in Advanced Machine Learning Systemscited by 258
22 more
  1. 2017DeepStack: Expert-Level Artificial Intelligence in No-Limit Pokercited by 233
  2. 2023AI Control: Improving Safety Despite Intentional Subversioncited by 229
  3. 2022Will we run out of data? Limits of LLM scaling based on human-generated datacited by 195
  4. 2018Categorizing Variants of Goodhart's Lawcited by 149
  5. 2023Taken out of context: On measuring situational awareness in LLMscited by 143
  6. 2022Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goalscited by 126
  7. 2019Optimal Policies Tend to Seek Powercited by 119
  8. 2021Consequences of Misaligned AIcited by 80
  9. 2018The Surprising Creativity of Digital Evolution: A Collection of Anecdotes from the Evolutionary Computation and Artificial Life Research Communitiescited by 78
  10. 2025Demonstrating specification gaming in reasoning modelscited by 56
  11. 2024Thousands of AI Authors on the Future of AIcited by 48
  12. 2025Evaluating Frontier Models for Stealth and Situational Awarenesscited by 37
  13. 2024Towards evaluations-based safety cases for AI schemingcited by 33
  14. 2024Mastering Board Games by External and Internal Planning with Language Modelscited by 31
  15. 2023Provably safe systems: the only path to controllable AGIcited by 29
  16. 2024The Shutdown Problem: An AI Engineering Puzzle for Decision Theoristscited by 22
  17. 2025Evaluation Faking: Unveiling Observer Effects in Safety Evaluation of Frontier AI Systemscited by 14
  18. 2023Game Theory with Simulation of Other Playerscited by 13
  19. 2022Circumventing interpretability: How to defeat mind-readerscited by 6
  20. 2024Game Theory with Simulation in the Presence of Unpredictable Randomisationcited by 3
  21. 2013Implementation without commitment in moral hazard environmentscited by 1
  22. 2023Should the Timing of Inspections be Predictable?cited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 1; 2026: 5.

6 papers.

  1. 2025Scheming Ability in LLM-to-LLM Strategic Interactionscited by 5
  2. 2026Discovering Differences in Strategic Behavior Between Humans and LLMscited by 3
  3. 2026Reward Shaping for (Inference-Time) Alignment: A Stackelberg Game Perspectivecited by 2
  4. 2026Decision Making under Imperfect Recall: Algorithms and Benchmarkscited by 0
  5. 2026Shutdown Safety Valves for Advanced AIcited by 0
  6. 2026An Investigation of the NeurIPS and ICML 2025 Position Trackscited by 0