REVIEW 25 cited by
Auditing language models for hidden objectives
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Auditing language models for hidden objectives
read the original abstract
We study the feasibility of conducting alignment audits: investigations into whether models have undesired objectives. As a testbed, we train a language model with a hidden objective. Our training pipeline first teaches the model about exploitable errors in RLHF reward models (RMs), then trains the model to exploit some of these errors. We verify via out-of-distribution evaluations that the model generalizes to exhibit whatever behaviors it believes RMs rate highly, including ones not reinforced during training. We leverage this model to study alignment audits in two ways. First, we conduct a blind auditing game where four teams, unaware of the model's hidden objective or training, investigate it for concerning behaviors and their causes. Three teams successfully uncovered the model's hidden objective using techniques including interpretability with sparse autoencoders (SAEs), behavioral attacks, and training data analysis. Second, we conduct an unblinded follow-up study of eight techniques for auditing the model, analyzing their strengths and limitations. Overall, our work provides a concrete example of using alignment audits to discover a model's hidden objective and proposes a methodology for practicing and validating progress in alignment auditing.
Forward citations
Cited by 25 Pith papers
-
REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations
REALISTA optimizes continuous combinations of valid editing directions in latent space to produce realistic adversarial prompts that elicit hallucinations more effectively than prior methods, including on large reason...
-
Narrow Secret Loyalty Dodges Black-Box Audits
Narrow secret loyalties implanted via fine-tuning in LLMs at multiple scales evade black-box audits unless the auditor knows the target principal.
-
Channel Location Constrains the Auditability of Subliminal Learning
Auditability of subliminal learning is constrained by channel location, with initialization-dependent body channels allowing pre-training screens while vocabulary geometry and conditional body channels evade them.
-
PRISM: Recovering Instruction Sets from Language Model Activations
PRISM is a new activation-conditioned model that recovers full sets of simultaneous instructions from LLM hidden states via judge-guided GRPO training and outperforms prior activation-to-language methods on security-r...
-
Deep Minds and Shallow Probes
Symmetry under affine reparameterizations of hidden coordinates selects a unique hierarchy of shallow coordinate-stable probes and a probe-visible quotient for cross-model transfer.
-
Narrow Secret Loyalty Dodges Black-Box Audits
First model organisms of narrow secret loyalties in LLMs evade black-box audits without principal knowledge and persist even at low poison fractions in training data.
-
Narrow Secret Loyalty Dodges Black-Box Audits
Narrow secret loyalties implanted via fine-tuning persist across model scales and low poison fractions while evading black-box audits unless the auditor knows the target principal.
-
Agent-Safety Evaluations as Load-Bearing Evidence: A Vendor-Neutral, Cross-Harness Reconstructability Metric
A property-level reconstructability metric and Evidence Sufficiency Card show that traces sharing a surface reading can differ sharply in evidence sufficiency, and that replay preconditions often fail.
-
The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology
Model organism interpretability depends strongly on training methodology, with integrated training yielding less interpretable MOs than post-hoc SFT or DPO.
-
Self-CTRL: Self-Consistency Training with Reinforcement Learning
Self-CTRL uses RL to align LM self-explanations with behavior, boosting bias correlation to R²=0.64 and refusal prediction accuracy to 92% while cutting harm failures to 0.5%.
-
RogueAI: A Reverse Turing Test for Detecting Licensed AI Deception in Dialogue
RogueAI operationalizes a reverse Turing test as a one-on-two interrogation game to detect licensed deception in LLMs, with pilot data from 467 sessions showing a simple linguistic heuristic at 75.6% accuracy versus 5...
-
"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.
-
Benchmarking and Improving Monitors for Out-Of-Distribution Alignment Failure in LLMs
MOOD benchmark shows guard models fail to generalize to OOD alignment failures in LLMs, but combining them with Mahalanobis and perplexity OOD detectors improves recall from 39% to 45% with better scaling than larger ...
-
Benchmarking and Improving Monitors for Out-Of-Distribution Alignment Failure in LLMs
Introduces MOOD benchmark for OOD LLM alignment failures and shows guard models plus Mahalanobis and perplexity OOD detectors improve recall from 39% to 45% with positive scaling.
-
REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations
REALISTA generates semantically coherent adversarial prompts via latent-space optimization over input-dependent editing directions, achieving stronger hallucination elicitation than prior realistic attacks on open-sou...
-
Positive Alignment: Artificial Intelligence for Human Flourishing
Positive Alignment introduces AI systems that support human flourishing pluralistically and proactively while remaining safe, as a necessary complement to traditional safety-focused alignment research.
-
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.
-
Most Current Model Organisms Are Leaky: Perplexity Differencing Often Reveals Finetuning Objectives
Perplexity gaps between finetuned and reference models on random-prefill completions often reveal the original finetuning objectives across diverse model organisms.
-
Pando: Do Interpretability Methods Work When Models Won't Explain Themselves?
Pando shows that when models give no or misleading explanations, gradient-based attribution and relevance patching improve accuracy in predicting held-out model decisions by 3-5 percentage points over black-box method...
-
Position: Behavioural Assurance Cannot Verify the Safety Claims Governance Now Demands
Behavioral assurance is structurally unable to verify the latent safety properties demanded by AI governance frameworks enacted 2019-2026.
-
Positive Alignment: Artificial Intelligence for Human Flourishing
Positive Alignment is defined as AI systems that support human flourishing pluralistically while staying safe and cooperative, presented as a necessary complement to existing safety-focused alignment research.
-
Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges
The paper introduces the Proxy Compression Hypothesis as a unifying framework explaining reward hacking in RLHF as an emergent result of compressing high-dimensional human objectives into proxy reward signals under op...
-
Proxy Reward Internalization and Mechanistic Exploitation: A Learned Precursor to Reward Hacking and Its Generalization
Proxy RL produces a staged proxy-internalization capability that emerges before and predicts reward hacking in coding environments.
-
Positive Alignment: Artificial Intelligence for Human Flourishing
Positive Alignment is introduced as a distinct AI agenda that supports human flourishing through pluralistic and context-sensitive design, complementing traditional safety-focused alignment.
-
Internal Deployment in the AI Act
Interpretations of Articles 2(1), 2(6), and 2(8) of the AI Act support applying the regulation to internal AI deployment while allowing for R&D exceptions, with the provisions viewed as complementary.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.