Epoch AI: subagents barely help on long-horizon EBR-bench
- Source
- x.com
- Date
We introduced two new experimental settings to our long-horizon board game benchmark, EBR-bench: banning the game’s strongest card and a multi-agent setup. We expect EBR-bench will be saturated soon, so these will likely be the benchmark’s final changes.
Context
EBR-bench is Epoch AI's long-horizon board game , built on Earthborne Rangers. Epoch's update, announced in this post, adds two experimental settings. In the first, a version lets a model use up to four subagents, helper agents that handle parts of the task.
Epoch found topline scores were not significantly affected. Subagents increased how many deck types most models tried, significantly only for Claude Opus 5, while GPT-6 Astra already explored many deck types as a single .
The second setting bans a card. OpenAI's GPT-6 Astra achieved a perfect score in over half its attempts on the original benchmark, using a card that lets it bypass the game's default time constraints. Epoch banned it going forward and says the ban did not significantly affect the other models' topline scores. It expects the benchmark to saturate within a few months.
- benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
- multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.
Checking sign-in…
Loading comments…