MolmoAct 2 Evaluation gallery
IN PROGRESS · PROVISIONAL Only fully completed, validated task shards are shown. Campaign coverage and aggregate rates can change.
Paper protocol Task 003 Validated task

open the top drawer and put the bowl inside

BDDL language: open the top layer of the drawer and put the bowl inside

Completed task success rate
98.0%
49 of 50
Control horizon
300
Maximum steps
Camera videos
100
Two per rollout
Evaluation time
7m 29s
Task process wall time

Canonical catalog

Task definition

Model instruction
open the top drawer and put the bowl inside
BDDL language
open the top layer of the drawer and put the bowl inside
Problem
libero_tabletop_manipulation
Objects
akita_black_bowl_1, cream_cheese_1, plate_1, wine_bottle_1
Fixtures
flat_stove_1, main_table, wine_rack_1, wooden_cabinet_1
Initial-state predicates
7
Fixed init states
50
Exact initial-state predicates 
(on wine_bottle_1 main_table_wine_bottle_region)
(on akita_black_bowl_1 main_table_akita_black_bowl_region)
(on plate_1 main_table_plate_region)
(on cream_cheese_1 main_table_cream_cheese_region)
(on wooden_cabinet_1 main_table_cabinet_region)
(on flat_stove_1 main_table_stove_region)
(on wine_rack_1 main_table_wine_rack_region)

Open exact BDDL definition

Simulator truth

Completion criteria

LIBERO checks these BDDL goal predicates after every action. All conditions must be true simultaneously in the same simulator step for the rollout to succeed.

  1. akita black bowl 1 is in wooden cabinet 1 top region

    (in akita_black_bowl_1 wooden_cabinet_1_top_region)
Exact parsed BDDL goal state 
(in akita_black_bowl_1 wooden_cabinet_1_top_region)

Selected rollout

LIBERO-Goal · Task 003 · Episode 1

Success
Agent view observation.images.image
Eye in hand observation.images.image2
Videos are not loaded yet.

Rollout record

Episode metadata

Link to episode
Episode
1
Seed
1000
Init state
0
Control steps
180
Wall time
9.8s
Outcome
Success

Evidence browser

All rollouts

Select one episode to mount only its two camera streams in the player.

Episode Outcome Seed Init state Steps Video
1 Success 1000 0 180 152 KB
2 Success 1001 2 183 152 KB
3 Success 1002 4 176 140 KB
4 Success 1003 6 175 147 KB
5 Success 1004 8 178 149 KB
6 Success 1005 10 180 152 KB
7 Success 1006 12 177 146 KB
8 Success 1007 14 178 151 KB
9 Success 1008 16 223 186 KB
10 Success 1009 18 220 200 KB
11 Success 1010 20 168 138 KB
12 Success 1011 22 178 155 KB
13 Success 1012 24 182 151 KB
14 Success 1013 26 171 154 KB
15 Success 1014 28 173 152 KB
16 Success 1015 30 183 149 KB
17 Success 1016 32 164 140 KB
18 Success 1017 34 178 147 KB
19 Success 1018 36 189 156 KB
20 Success 1019 38 179 148 KB
21 Success 1020 40 170 152 KB
22 Success 1021 42 172 145 KB
23 Success 1022 44 174 150 KB
24 Success 1023 46 173 146 KB
25 Success 1024 48 180 152 KB
26 Success 1025 0 173 150 KB
27 Success 1026 2 177 152 KB
28 Success 1027 4 169 138 KB
29 Success 1028 6 176 145 KB
30 Success 1029 8 179 152 KB
31 Success 1030 10 166 152 KB
32 Success 1031 12 169 141 KB
33 Success 1032 14 168 141 KB
34 Success 1033 16 239 196 KB
35 Success 1034 18 172 151 KB
36 Failure 1035 20 300 110 KB
37 Success 1036 21 188 158 KB
38 Success 1037 23 178 147 KB
39 Success 1038 25 174 146 KB
40 Success 1039 27 185 159 KB
41 Success 1040 29 165 151 KB
42 Success 1041 31 170 151 KB
43 Success 1042 33 174 149 KB
44 Success 1043 35 178 157 KB
45 Success 1044 37 177 151 KB
46 Success 1045 39 191 150 KB
47 Success 1046 41 190 150 KB
48 Success 1047 43 177 145 KB
49 Success 1048 45 170 145 KB
50 Success 1049 47 177 153 KB

Integrity

Task attestation

Status
complete
Completed
Aug 16, 2026, 1:38 AM UTC
Manifest SHA-256
04721c9b9e…015b80a5
Aggregate SHA-256
25a6a2b7e4…e0b412d9
Episode index SHA-256
cf065d35ba…e3b21f47
Validation SHA-256
35a9f3d64d…c765a9ba
Artifact inventory SHA-256
316af0fe59…cc1469ba