Perplexity reports a potentially useful but not yet independently verified improvement in computer-agent reliability from failure-focused post-training.
Limits of the evidence: The item is an independent report of a Perplexity study, provides only a brief summary, and does not include independent replication or full experimental details.
What to watch
Taken from the first report; the whole story has not been re-assessed yet.
Perplexity Trains Its Computer Agent on Real Mistakes With Hint-Guided Self-Distillation
MarkTechPostSep 25, 2026First report
No revisions yet.
No maturity ladder applies to this desk.
Sens.ai aggregates and assesses published reporting. The assessment above is machine generated; the original sources are authoritative.