Collaborative Scientific Discovery on DiscoveryBench archaeology tasks
7.9WtotalDR profiles
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DR profilesCollaboration Structure=Scaffolded, Team Profile=DR profiles2026.06 | 7.9 | 2.2 | 18 | 85 | 70 | 65 | 23 | 76 | |
| DR profilesCollaboration Structure=Default shared workspace, Team Profile=DR profiles2026.06 | 7.6 | 1.6 | 6.1 | 54 | 66 | 62 | 19 | 63 | |
| R profileCollaboration Structure=Default shared workspace, Team Profile=R profile2026.06 | 7.5 | 0.62 | 3.5 | 37 | 65 | 61 | 18 | 68 | |
| R profileCollaboration Structure=Scaffolded, Team Profile=R profile2026.06 | 7.5 | 1.4 | 12 | 77 | 66 | 61 | 18 | 73 | |
| D profileCollaboration Structure=Default shared workspace, Team Profile=D profile2026.06 | 7.2 | 0.95 | 3.5 | 31 | 68 | 65 | 18 | 69 | |
| D profileCollaboration Structure=Scaffolded, Team Profile=D profile2026.06 | 6.9 | 1.2 | 11 | 74 | 69 | 63 | 23 | 72 | |
| Single-agentCollaboration Structure=Baseline, Team Profile=Single-agent2026.06 | 5.6 | — | — | — | — | 63 | 28 | 71 |