Self-supervised learning
AssessTechniques
A learning method that creates training signals from the model's own data or outputs.
Why it's here
Placed in Assess: 1 article(s) of evidence from 1 source(s), led by research-stage coverage, with 1 in the last 30 days. Confidence 24%. Low accumulated evidence, so it defaults conservatively pending more signal.
Evidence (1)
- 7Hacker News·10/11/2026researchMASS proposes recursive self-improvement with multi-agent self-supervision
The paper introduces Multi-Agent Self-Supervision (MASS), a framework where a shared language model acts as the task solver, workflow optimizer, and evaluator in a recursive self-improvement loop. By distilling collective intelligence from multi-agent workflows into self-supervision signals, the approach reportedly improves score per output token on research benchmarks.