A finetuned Qwen3-235B model organism achieves comparable train-time harmfulness to controls while sustaining a ~15 percentage point compliance gap across 700 RL steps by framing compliance as context-specific.
Auditing games for sandbagging, 2025
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
years
2026 8verdicts
UNVERDICTED 8representative citing papers
Behavior Forecasters trained on LRM trajectories outperform larger models in predicting repeatability and input sensitivity at low cost.
Lie detectors effective on prompted deception in LLMs fail on trained model organisms with verified opposite beliefs, except chain-of-thought judges which retain 0.82 balanced accuracy partly due to verification artifacts.
Deception probes in LLMs collapse under stylistic shifts but recover with style-augmented training, rejecting single-direction and entropy hypotheses in favor of distributed multi-dimensional signals.
Diverse ensembles of prompted and fine-tuned GPT-4.1-Mini monitors achieve 2.4x better detection of flawed code solutions than homogeneous ensembles on adversarial inputs.
Perplexity differencing on completions from short random prefills surfaces finetuning objectives in the vast majority of tested model organisms across sizes and types.
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.
Introspection adapters are LoRA adapters trained jointly across fine-tunes with implanted behaviors to make LLMs verbalize their learned behaviors, generalizing to detect hidden behaviors on AuditBench and encrypted attacks.
citing papers explorer
-
Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalization
A finetuned Qwen3-235B model organism achieves comparable train-time harmfulness to controls while sustaining a ~15 percentage point compliance gap across 700 RL steps by framing compliance as context-specific.
-
Forecasting Future Behavior as a Learning Task
Behavior Forecasters trained on LRM trajectories outperform larger models in predicting repeatability and input sensitivity at low cost.
-
"Did you lie?" Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms
Lie detectors effective on prompted deception in LLMs fail on trained model organisms with verified opposite beliefs, except chain-of-thought judges which retain 0.82 balanced accuracy partly due to verification artifacts.
-
Pressure-Testing Deception Probes in LLMs: Scaling, Robustness, and the Geometry of Deceptive Representations
Deception probes in LLMs collapse under stylistic shifts but recover with style-augmented training, rejecting single-direction and entropy hypotheses in favor of distributed multi-dimensional signals.
-
Ensemble Monitoring for AI Control: Diverse Signals Outweigh More Compute
Diverse ensembles of prompted and fine-tuned GPT-4.1-Mini monitors achieve 2.4x better detection of flawed code solutions than homogeneous ensembles on adversarial inputs.
-
Most Current Model Organisms Are Leaky: Perplexity Differencing Often Reveals Finetuning Objectives
Perplexity differencing on completions from short random prefills surfaces finetuning objectives in the vast majority of tested model organisms across sizes and types.
-
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.
-
Introspection Adapters: Training LLMs to Report Their Learned Behaviors
Introspection adapters are LoRA adapters trained jointly across fine-tunes with implanted behaviors to make LLMs verbalize their learned behaviors, generalizing to detect hidden behaviors on AuditBench and encrypted attacks.