Can AI agents predict whether they will succeed at a task? We study agentic uncertainty by eliciting success probability estimates before, during, and after task execution. All results exhibit agentic overconfidence: some agents that succeed only 22% of the time predict 77% success. Counterintuitively, pre-execution assessment with strictly less information tends to yield better discrimination than standard post-execution review, though differences are not always significant. Adversarial prompting reframing assessment as bug-finding achieves the best calibration.
Think Miniml / Research
Agentic Uncertainty Reveals Agentic Overconfidence
Can AI agents predict whether they will succeed at a task? We study agentic uncertainty by eliciting success probability estimates before, during, ...
Full paper · available on arxiv.org
Read paper
More from Think Miniml
Keep reading.
Research
Logit-Contribution Scoring Identifies Non-Literal Retrieval Heads
In long-context use, large language models frequently synthesize answers from the meaning of a relevant context span rather than literally copy-pasting them.…
Read more →
Research
Human-LLM Dialogue Improves Diagnostic Accuracy in Emergency Care
Clinical decision-making in emergency medicine demands rapid, accurate diagnoses under uncertainty. Despite benchmark progress, evidence for LLMs as interactive aids in live…
Read more →
Research
SCOPE: Self-Play via Co-Evolving Policies for Open-Ended Tasks
Self-play can train language models without external supervision. However, existing methods require rule-checkable answers, leaving open-ended tasks dependent on curated prompts or…
Read more →Start the conversation
Talk to a senior consultant.
30 minutes. Bring a problem you’re stuck on — we’ll tell you what we’d do next.
Book a consultation →