Skip to content
Robotensor

One-demo in-context imitation

Duel

Crown held5 Sept 2026, 17:20 UTCsmoke duela0e9d3e8b5…740d1f

Reigning model

Holds the crown

robotensor/bpp-libero-genesis

Revision
478bbecdfd0ce3d66d49933319e7e9f2e874c9bd
Submission key
dfc33596f6aad135

37.5%

Average of the four axes

Spatial change
Environment change
Object change
Composition

Challenger

robotensor/bpp-libero-genesis

Revision
db9c478e2f9e7e91333994d66d3ec0dfdcebb874
Submission key
c381e60f7bef11a7

37.5%

Average of the four axes

Spatial change
Environment change
Object change
Composition
Margin check37.5% − 37.5% = 0.0 ptsneeds ≥ +3.0 ptsNot met

Verdict derivation

recomputed from 8 unitsCrown held

The ledger

Per-axis success rates for both sides and their difference
AxisReigningChallengerΔ
Spatial change50.0%1 / 250.0%1 / 20.0 pts
Environment change50.0%1 / 250.0%1 / 20.0 pts
Object change0.0%0 / 20.0%0 / 20.0 pts
Composition50.0%1 / 250.0%1 / 20.0 pts
Averageneeds ≥ 40.5%37.5%37.5%0.0 pts

challenger average ≥ reigning average + +3.0 pts

37.5% ≥ 37.5% + +3.0 pts · needs ≥ 40.5%

Not met

The reigning model kept the crown

37.5% vs 37.5% · 0.0 pts

Paired outcomes

diagnostic, does not decide the crown

Won

0

Lost

0

Tied

8

Decided

0

Void

0

The paired record

8 units · 0 decided · 8 tied

  • Reigning
    50.0%1/2
    Challenger
    50.0%1/2
    Δ
    0.0 pts
  • Reigning
    50.0%1/2
    Challenger
    50.0%1/2
    Δ
    0.0 pts
  • Reigning
    0.0%0/2
    Challenger
    0.0%0/2
    Δ
    0.0 pts
  • Reigning
    50.0%1/2
    Challenger
    50.0%1/2
    Δ
    0.0 pts
  • Reigning
    100.0%1/1
    Challenger
    100.0%1/1
  • Tied
    • chain
    • source split libero_10
    Reigning
    Pass
    Challenger
    Pass
  • Reigning
    0.0%0/1
    Challenger
    0.0%0/1

Provenance

Record
a0e9d3e8b500…17740d1f
Track
One-demo in-context imitation
Kind
Duel
Finished
5 Sept 2026, 17:20 UTC
Wall time
4m 59s
Duel size
smoke
Units per axis
2 × 4 axes = 8 per side
Score margin in force
+3.0 pts
Duel id
8543f53ce9ff…4a6e28db
Pool id
7bf31989bda4…8847d259
Protocol version
1
Protocol fingerprint
908bcc0eb92f…ec8dfeca
Cycle
2
Media attached
16 clips