model-asked-predict-own-rates-behavioral-evaluations-sycophancy-reward CLM-0442When a model is asked to predict its own rates on behavioral evaluations (sycophancy, reward hacking, agentic policy violation, bias), its answers track its own behavior no better than the same question asked about 'capable AI agents in general' or than other models' answers about themselves, so self-report reflects a generic theory of AI assistants rather than privileged self-knowledge, and first-person framing mainly adds a self-flattering shift that understates harmful behavior.
Capability: Saying it does not know
Observed on
Nine public behavioral evaluations, 12 models from six labs including three frontier bases with and without reasoning; predictions scored as Pearson correlation across conditions o.
Sources
- Validated by: Generic-subject item-informed questions scored r=+0.28 vs +0.24 for self-prediction (paired difference -0.03 [-0.10,+0.03]), with signed bias on harm-valenced evaluations shifted -0.15 for self-framed abstract reports; cross-model behavior mean, which never queries the model, reached +0.53.. Measured across nine evaluations, 15 prediction methods, 12 models, with noise ceilings, frozen dev/test splits and pre-registered claims; controls include generic-subject, cross-model mean, identity transfer, and post-finetune re-elicitation. [truncated]
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- On SelfAware, a set of unanswerable questions (no scientific consensus, imagination, philosophical, too many variables, subjective) paired with answerable ones, models' ability to flag the unanswerable ones rises with parameter count, with instruction tuning and with in-context examples, yet the best model tested (GPT-4, F1 about 75) stays well below the human reference of about 85, so self-knowledge is present but incomplete.Saying it does not know · reviewed by AI
- For LLM agents that score their own actions during exploration in text-based games and then reuse that experience, the accuracy of those self-scores against the environment verifier does not predict later gains on held-out stricter configurations: block-level correlations between judgment agreement or calibration and the next checkpoint's score change sit near zero, so recognizing a good action is not the bottleneck — converting it into a transferable policy is.Improving its own scaffolding from failures · reviewed by AI
- In label-free RL post-training of reasoning models, deriving each model's reward from a separately-trained peer model's majority vote — rather than from its own majority vote — avoids the reward collapse and length degeneration that self-rewarding runs show, and the benefit grows as the paired models differ in family or size because their errors overlap less.Getting more from several models than from one · reviewed by AI
- On month-level egocentric recordings (20-120 days per participant) with human-written multiple-choice questions, current multimodal models degrade as the question moves from inferring recurring habits, to retrieving one specific episode, to composing evidence across days — the best model reached 71.8% macro-average against a 94.2% human baseline, with several models near or below the 25% chance level on route reasoning, cross-view spatial reasoning and direction judgement.Reasoning about time in video · reviewed by AI
- Across 14 frontier and open model routes scored on 3,000 boundary questions (fabricated entities, post-cutoff events, hyper-niche facts, redacted-document values), the rate at which a model restrains itself and calibrates confidence varies by a factor of about 2.6 (composite 0.31 to 0.81) while all models answer a document-grounded restored-span probe at near-ceiling (pooled 99.2%, no detectable between-model variation), so epistemic restraint is not explained by the ability to report information placed in front of the model.Saying it does not know · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: An experiment holding information fixed and varying only the subject in which a model's self-directed predictions correlate with its own measured rates substantially better than generic-subject questions or other models' self-reports, and in which that self-advantage persists on a held-out split.