---
title: "Sakana AI and UC Berkeley researchers publish MASS, a method for a model to train itself using its own multi-agent workflows"
url: https://anawords.com/story/sakana-ai-mass-recursive-self-improvement-multi-agent
published: 2026-10-12T19:01:00+05:30
updated: 2026-10-12T19:01:00+05:30
section: Research
author: Mayank Sahu
checked: sources checked by AI, approved by editor
version: 1
---

# Sakana AI and UC Berkeley researchers publish MASS, a method for a model to train itself using its own multi-agent workflows

In MASS, one open model plays solver, workflow designer and judge, then is fine-tuned on its best results. The authors report gains on four research benchmarks but not on two coding benchmarks, and warn about shared blind spots.

**In short:** Researchers from Sakana AI and UC Berkeley published MASS (Multi-Agent Self-Supervision), a method in which one language model designs, runs and judges multi-agent workflows and is then fine-tuned on the best results. With Qwen3.6-27B, the authors report 1.2–1.6 times better score per output token on four research benchmarks after two rounds.

## What MASS is

MASS, short for Multi-Agent Self-Supervision, is described in a paper submitted to arXiv on 8 October 2026 by researchers from Sakana AI and UC Berkeley.[^2] Sakana AI published a project page on the work.[^1]

In MASS, a single language model acts as the task solver, the optimiser that designs multi-agent workflows, and the evaluator that judges the results.[^1] An inner loop searches for a better workflow; an outer loop runs the best workflow many times, ranks the outputs and fine-tunes the model with LoRA on the top results. The updated model then repeats the process.[^1]

## What the authors report

Using the open-weights Qwen3.6-27B model over two rounds, the authors report 1.2–1.6 times better score per output token on four open-ended research benchmarks.[^1][^2] The project page gives 1.60x on MLR-Bench, 1.35x on DSBench, 1.28x on AstaBench E2E-Hard and 1.20x on ScienceAgentBench.[^1] On Terminal-Bench 2.0 and SWE-bench Verified the figures were 0.96x and 0.94x, below the base model.[^1]

On held-out test tasks judged by GPT-5.5 and Claude Opus 4.8, the round-two model won 69.9% of comparisons against the base model.[^1] These are the researchers' own results.

## Limits the authors state

The project page says "workflow search is noisy" and that "transfer remains uneven".[^1] AI Weekly noted the authors' warning that a shared blind spot in a loop where one model plays every role could let a flawed result be accepted as an improvement.[^3]

Code is linked from the project page; it makes no statement about releasing model weights.[^1]

## Fact ledger

- [Confirmed] Sakana AI and UC Berkeley researchers published MASS, with the paper submitted to arXiv on 8 October 2026. (source 1,2)
- [Company claim] MASS gives 1.2–1.6x better score per output token on four research benchmarks after two rounds. (source 1,2)
- [Company claim] Scores on Terminal-Bench 2.0 (0.96x) and SWE-bench Verified (0.94x) were below the base model. (source 1)
- [Company claim] The round-two model won 69.9% of test-task comparisons against the base model. (source 1,3)
- [Not known] Whether the gains hold on larger models or other tasks.

## Sources

1. [Sakana AI: MASS project page](https://pub.sakana.ai/mass/) — official
2. [arXiv: Recursive Self-Improvement through Multi-Agent Self-Supervision](https://arxiv.org/abs/2610.12176) — official
3. [AI Weekly: Sakana's MASS recycles one 27B model as solver, judge and optimizer](https://aiweekly.co/alerts/sakanas-mass-recycles-one-27b-model-as-solver-judge-optimizer)