aiIMAGE
Perplexity cuts tool-call failures 21% with self-distillation
Perplexity published research on hint-guided self-distillation that teaches its Computer model from its own errors, cutting tool-call failures by 21.2% relative in a live A/B test.
Perplexity
@perplexity_aiArchived 2026.09.24Originally posted 2026.09.23
Perplexity @perplexity_aiOriginally posted 2026.09.23
New research: We post-trained a Computer model to learn from its own errors using hint-guided self-distillation. In a live A/B test, a later trained checkpoint reduced tool-call failures by 21.2% relative to an earlier checkpoint. https://t.co/3MFrp1yxDt
x.com/perplexity_ai/status/2102493613192298913Archived 2026.09.24 01:48