SPRIIA study of persistent information
Generality · Formation

Overcooked

What persists can be a partner.

Across distinct cooperative episodes, a fixed partner identity supplies a relation for learning behavioral context.

Partner identityacross disjoint episodes
Motion preview
Overcooked environment replayRendering: JaxMARL contributors · illustrative controller rollout. Source & changes
Across independent experiences

What persists.

The fixed neural partner identity.

Across distinct realizations

What changes.

Disjoint episodes, positions, actions and outcomes.

Experience → prediction

The task.

Prior episodes + current query.

→ Ego-action prediction.

Evaluated instance

The comparison.

Native multi-episode AD action learner · Align only.

Controls. Matched history slot + variance/covariance regularization.

Interpreting the evidence

What stays fixed in the test.

The grounded_coord_simple layout and partner relation are explicit. A100 and VC100 share the matched training/evaluation recipe; history budgets are analyzed separately.

What this setting establishes.

The application measures partner-behavior readout, then cooperation across history budgets. The completed three-seed record separates familiar from held-out-development partners.

Scope of the evidence

Held-out-development cooperation is not a stable positive headline: A100−VC100 is −1.111 ± 11.437 points across source seeds. The displayed native scene is not a trained-policy success demonstration.

Full result record

Measurements and comparisons.

Reported results are kept with their own populations and conditions. Development, held-out and sealed comparisons remain separate.

01 / Study Table 40

Align − H+VC: partner readout and cooperation differences

3 terminal source seeds 4200/4201/4202; familiar and held-out-development partners separated

Paired Align minus H+VC differences: negative probe-MSE changes and positive return changes are favorable.

Align − H+VC: partner readout and cooperation differences
HistoryΔ familiar probe MSEΔ familiar returnΔ held-out dev. return
25%-0.00567 ± 0.008101.489 ± 2.335-2.222 ± 15.207
50%-0.01344 ± 0.007280.389 ± 1.11018.444 ± 20.659
100%-0.02120 ± 0.006101.889 ± 2.175-1.111 ± 11.437

Bold pairs full-history familiar and held-out-development return effects.

  • 100%The familiar effect is positive in all three seeds; held-out-development effects have mixed signs. ± is sample SD, not a confidence interval.

Paired Align-minus-H+VC differences are means ± sample SD across three training seeds. Return averages episodes 6–20 per partner. Full-history familiar effects are positive in every seed; held-out-development effects have mixed signs. The two partner populations remain separate.

02 / Study Table 41

Partner behavior and cooperation across history budgets — A. Behavioral readout and representation

3 terminal source seeds 4200/4201/4202; familiar and held-out-development partners separated

Behavioral probe MSE ↓; distance ratio; cooperation return ↑

Partner behavior and cooperation across history budgets — A. Behavioral readout and representation
HistoryRecipeFamiliar MSE ↓Held-out dev. MSE ↓Distance ratio
25%VC0.06788 ± 0.006670.13883 ± 0.042301.343 ± 0.061
25%A0.06220 ± 0.009740.09733 ± 0.013841.917 ± 0.071
50%VC0.06934 ± 0.008110.14091 ± 0.023241.408 ± 0.019
50%A0.05590 ± 0.007360.08966 ± 0.012402.329 ± 0.326
100%VC0.07516 ± 0.007910.12395 ± 0.028571.458 ± 0.036
100%A0.05395 ± 0.011060.10631 ± 0.010462.269 ± 0.206
100%Random0.07978 ± 0.014230.14836 ± 0.032031.460 ± 0.066

Bold marks Align’s behavioral-probe errors within each history budget.

Dashed rules separate budgets and the probe-only Random control.

  • 100% · RandomRandom has probe coverage only. Familiar and held-out-development probe populations stay separate.

Means ± sample SD across three terminal checkpoints. A is Align; VC is H+VC. Familiar and held-out-development probe populations remain separate. Distance ratios use the familiar behavioral-signature bank; Random has probes only, without return evaluation.

03 / Study Table 41

Partner behavior and cooperation across history budgets — B. Cooperation returns

3 terminal source seeds 4200/4201/4202; familiar and held-out-development partners separated

Behavioral probe MSE ↓; distance ratio; cooperation return ↑

Partner behavior and cooperation across history budgets — B. Cooperation returns
HistoryRecipeFamiliar partnersHeld-out dev. partners
25%VC15.778 ± 1.262-1.333 ± 4.372
25%A17.267 ± 1.768-3.556 ± 14.852
50%VC15.200 ± 0.200-19.556 ± 2.341
50%A15.589 ± 0.916-1.111 ± 18.407
100%VC14.044 ± 1.500-19.111 ± 8.572
100%A15.933 ± 0.742-20.222 ± 3.672

Bold keeps both full-history policies and partner populations in view.

Dashed rules separate offline history budgets.

  • 100% · AReturns come from frozen-policy episodes with no optimizer updates. The held-out-development population is not a stable positive generalization result.

Means ± sample SD across three training seeds. Frozen policies run without optimizer updates; return averages episodes 6–20 within each partner, then partners within each cohort. The full-history held-out-development Align-minus-H+VC difference is −1.111 ± 11.437, with mixed seed-level effects.