Brazilian researchers compare technology stacks for AI agents
A NIC.br and PUC-SP study tested three models, three protocols and four frameworks on 300 Mind2Web tasks.
Event / data period: 2026-10-09
Published: · Updated:
NIC.br and PUC-SP researchers published a study on 9 October introducing MASTIF, a framework for comparing technology combinations used in multi-agent systems. The evaluation crossed three language models, three communication protocols and four implementation frameworks across 300 tasks from the Mind2Web benchmark.
The paper reports that technology combinations differed across measures including task completion, time, concision, tool use and stability of social values; no single stack ranked best on every measure. The authors note limits in the task set and models evaluated. These results provide an experimental baseline, not a guarantee about agents in real-world use.
Data
Study scope
300 Mind2Web tasks; MASTIF experiment sample
| Tasks evaluated from the Mind2Web benchmark | 300 | 2026-10-09 |
Methodology
Count reported by the paper for its experimental scope; not a national statistic or field result.
Updated:
Subjects and places
Subject proposals
- Artificial intelligenceSuggested
- Subject suggested from the title or summary: AIarxiv.org ↗
Geographic references
Only source-provided geographic references are shown. Opening an external map is optional. Optional story maps use OpenStreetMap. Opening a map sends your IP address and this site’s address to the map provider; location data remains readable as text without loading the map.
- Brazil · Place role: Coverage · Place type: Country or area · Location precision: RegionalEvidence
- Explicit preserved editorial country/region; no city or national centroid inferredarxiv.org ↗