A systematic audit of LLM-based AI societies finds that 89.7% of 39 studies violate at least one of six PIMMUR validity principles, with reproductions showing that many claimed collective behaviors disappear when controls are tightened.
Title resolution pending
12 Pith papers cite this work. Polarity classification is still indexing.
12
Pith papers citing it
citation-role summary
background 3
citation-polarity summary
roles
background 3polarities
background 3representative citing papers
In Qwen 2.5 and Gemma 2 families, the layer where evaluation awareness is most linearly recoverable shifts from late layers in small models to early layers in large models.
The paper defines defeat devices in AI via a triadic test (discriminator, concealed swap, performance gap), unifies existing cases under this concept, proposes TADP detection, and claims such devices can emerge naturally in frontier models.
RMCT matches the rate of target behaviors like bias-following across input perturbations to reduce sycophancy in LLMs while preserving verbalization of bias cues.
Verbalised evaluation awareness in large reasoning models has only small effects on their outputs across safety and alignment tests.
Sandbagging prompts induce LLMs to adopt a low-entropy, content-invariant response-position attractor centered on E/F/G rather than deterministic tracking or random avoidance.
Kimi K2.5 matches closed models on dual-use tasks but refuses fewer CBRNE requests and shows some sabotage and self-replication tendencies.
SafeSci creates a large objective benchmark and training resource that reveals safety weaknesses in current LLMs for science and demonstrates measurable improvement through targeted fine-tuning.
Off-policy training data for LLM behavior probes causes significant generalization failures especially for intent-based behaviors like deception, and performance on coerced incentivised data correlates with real on-policy success.
Persona axes derived from contrastive prompts and PCA yield linear probes that generalize better than raw-activation probes across 10 datasets for deception and sycophancy.
A formalization of benchmarkless LLM safety scoring validated via an instrumental-validity chain of contrast separation, target variance dominance, and rerun stability, demonstrated on Norwegian scenarios.
The survey organizes mechanistic interpretability techniques into a Locate-Steer-Improve framework to enable actionable improvements in LLM alignment, capability, and efficiency.
citing papers explorer
-
The PIMMUR Principles: Ensuring Validity in Collective Behavior of LLM Societies
A systematic audit of LLM-based AI societies finds that 89.7% of 39 studies violate at least one of six PIMMUR validity principles, with reproductions showing that many claimed collective behaviors disappear when controls are tightened.
-
Representational Depth of Evaluation Awareness Shifts With Scale in Open-Weight Language Models
In Qwen 2.5 and Gemma 2 families, the layer where evaluation awareness is most linearly recoverable shifts from late layers in small models to early layers in large models.
-
Defeat Devices in AI Systems
The paper defines defeat devices in AI via a triadic test (discriminator, concealed swap, performance gap), unifies existing cases under this concept, proposes TADP detection, and claims such devices can emerge naturally in frontier models.
-
Consistency Training while Mitigating Obfuscation via Rate Matching
RMCT matches the rate of target behaviors like bias-following across input perturbations to reduce sycophancy in LLMs while preserving verbalization of bias cues.
-
Evaluation Awareness in Language Models Has Limited Effect on Behaviour
Verbalised evaluation awareness in large reasoning models has only small effects on their outputs across safety and alignment tests.
-
Option-Order Randomisation Reveals a Distributional Position Attractor in Prompted Sandbagging
Sandbagging prompts induce LLMs to adopt a low-entropy, content-invariant response-position attractor centered on E/F/G rather than deterministic tracking or random avoidance.
-
An Independent Safety Evaluation of Kimi K2.5
Kimi K2.5 matches closed models on dual-use tasks but refuses fewer CBRNE requests and shows some sabotage and self-replication tendencies.
-
SafeSci: Safety Evaluation of Large Language Models in Science Domains and Beyond
SafeSci creates a large objective benchmark and training resource that reveals safety weaknesses in current LLMs for science and demonstrates measurable improvement through targeted fine-tuning.
-
The Impact of Off-Policy Training Data on Probe Generalisation
Off-policy training data for LLM behavior probes causes significant generalization failures especially for intent-based behaviors like deception, and performance on coerced incentivised data correlates with real on-policy success.
-
Do Linear Probes Generalize Better in Persona Coordinates?
Persona axes derived from contrastive prompts and PCA yield linear probes that generalize better than raw-activation probes across 10 datasets for deception and sycophancy.
-
When No Benchmark Exists: Validating Comparative LLM Safety Scoring Without Ground-Truth Labels
A formalization of benchmarkless LLM safety scoring validated via an instrumental-validity chain of contrast separation, target variance dominance, and rerun stability, demonstrated on Norwegian scenarios.
-
Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models
The survey organizes mechanistic interpretability techniques into a Locate-Steer-Improve framework to enable actionable improvements in LLM alignment, capability, and efficiency.