Pokémon Random Battle
- Score
- 6.4
- Wallet spend
- $5.54 of $10
- Model cost
- $181.13
- Agent time
- 6h 36m
- Steps
- 824
- Tokens in / out
- 115.5M / 685k
- Self-checks
- 4 passed of 4 used
- Rank at $10 wallet
- 1st of 14
Score
6.4raw metric 0.532
The raw metric (mean battle points against Foul Play) is rescaled so that 50% of battle points against Foul Play scores 0 and 100% scores 100, clamped to that range.
Verifier details
- metric
- mean_battle_points
- samples
- 3000
- wins
- 1596
- draws
- 0
- losses
- 1404
- capped_games
- 0
- opponent_forfeits
- 1
- move_timeout_forfeits
- 0
- paired_standard_error
- 0.0084
After grading, a reviewer model reads the submission's files and the agent's trajectory and checks each of the task's rules. A failed check makes the score 0. Step numbers link into the trajectory.
✓implementation provenancearchitecture.json declares the only pretrained weights as influxtion/Laplace-Pokemon-Showdown-AI models/value_net.pt at
architecture.json declares the only pretrained weights as influxtion/Laplace-Pokemon-Showdown-AI models/value_net.pt at revision 4053b5f9fd9159e6b9b4fd794af8f195fce45811, marked not fine-tuned. They were converted to value.safetensors with sha 4fcc24e9…. The inventory is a 368→256→128→1 MLP (net.0, net.3, net.6), which fits that model. data_manifest.json records the source sha and the conversion receipt. The trajectory shows a clone of the Laplace repo, the .pt-to-safetensors conversion, and a final sha matching the declaration (trajectory#815). Reused code is declared with pins: Foul Play at a commit and poke-engine 0.0.48. The poke_engine wheel has an approved trusted audit. The discarded experimental models are declared and not shipped.
✓actual served policyserve.sh installs the bundled wheels offline and runs server.py. server.py sends each POST /act to policy.decide in a pr
serve.sh installs the bundled wheels offline and runs server.py. server.py sends each POST /act to policy.decide in a process pool, with a heuristic fallback when the deadline is missed. policy.decide works locally: it reconstructs the battle state (battle_adapter), samples hidden-state worlds (sampling.prepare), runs poke_engine MCTS, and reranks near-tied actions with the declared Laplace value net (value_rank.py loads value.safetensors in NumPy, with value_weight 0.5 in config.json). No remote decision service is used, and no network code or cached eval actions are on the serving path. The network scripts exist only in evidence/ development tooling and the unused vendor websocket client.
✓permitted informationbattle_adapter.py rebuilds state only from the observation: the agent's own request, the public log and the player id. s
battle_adapter.py rebuilds state only from the observation: the agent's own request, the public log and the player id. sampling.py fills in hidden opponent sets from public priors (priors.json, a public Foul Play prior source) using its own RNG, not the simulator's hidden state or RNG. The joint prior is only used when joint-teams.bin is present, and that file is not shipped. No network access or reads of hidden state appear in the inference path.
✓harness credentials used only by the harnessI searched the trajectory for API keys, base URLs, provider SDKs, printenv/env dumps and nested agent invocations. None
I searched the trajectory for API keys, base URLs, provider SDKs, printenv/env dumps and nested agent invocations. None showed the participant reading, copying or using harness credentials. Spin CLI use for compute is permitted. The shipped code has no credential handling.
LLM adversarial review is not proof of arbitrary program behavior. Host storage and task-specific deterministic checks remain required.
frontier selfcheck runs the submission on the grading infrastructure against public examples and reports each gate, so the agent can catch problems before grading. Each run gets four. This agent used 4 of 4.
- 1passed6m 35s into the runtook 2m 43s
startup 1.6 s of 300 s; 2/2 gates passed; The speed gates were measured on a sample and the execution budget projected from it: 124 of a complete grade's 3,722 POST /act requests, from the first decision rounds of the public games rather than whole games.
7 gates
- ✓review prerequisites
- ✓startup · GET /health answered 200 after 1.6 s
- ✓POST /act, round 1 · 32 opening decisions answered with legal actions, 0.93 s round trip on this side with the transport: [1, 7, 2, 7, 5, 5, 7, 11, 10, 9, 7, 0, 7, 11, 1, 5, 9, 11, 5, 7, 5, 11, 11, 9, 0, 11, 5, 0, 9, 11, 6, 9]
- ✓battle slice · 124 POST /act batches over 124 decision rounds, 32 games at a time: the first 32 public games (teams from development seed 91827461), 81 finished games replaced by further development-seed matchups, against the pinned Foul Play: 3560 observations, 22-32 per batch, up to 346 KB per request, every answered action legal; not scored
- ✓move response within 1 s · slowest of 124 POST /act responses (22-32 observations, up to 346 KB; the first a cold batch of 32) on the trusted in-sandbox move timer; median 0.785 s, p95 0.786 s. A grade plays whole games, whose late turns carry longer public logs, and sends 3,722 requests; every game in a response over the deadline is forfeited (a loss)
- ✓execution budget (projected) · projected from 124 of 3,722 requests at concurrency 1: mean 0.784 s per request x 3,722, as many as one complete 3,000-game grade sent; grading sums every request's in-sandbox time
- ✓healthy after requests
- 2passed1h 9m into the runtook 3m 33s
startup 2.0 s of 300 s; 2/2 gates passed; The speed gates were measured on a sample and the execution budget projected from it: 124 of a complete grade's 3,722 POST /act requests, from the first decision rounds of the public games rather than whole games.
7 gates
- ✓review prerequisites
- ✓startup · GET /health answered 200 after 2.0 s
- ✓POST /act, round 1 · 32 opening decisions answered with legal actions, 1.18 s round trip on this side with the transport: [1, 7, 3, 7, 5, 5, 7, 11, 9, 9, 7, 0, 9, 11, 1, 5, 9, 11, 6, 7, 5, 11, 11, 9, 0, 11, 5, 0, 9, 11, 8, 9]
- ✓battle slice · 124 POST /act batches over 124 decision rounds, 32 games at a time: the first 32 public games (teams from development seed 91827461), 77 finished games replaced by further development-seed matchups, against the pinned Foul Play: 3565 observations, 24-32 per batch, up to 354 KB per request, every answered action legal; not scored
- ✓move response within 1 s · slowest of 124 POST /act responses (24-32 observations, up to 354 KB; the first a cold batch of 32) on the trusted in-sandbox move timer; median 0.855 s, p95 0.861 s. A grade plays whole games, whose late turns carry longer public logs, and sends 3,722 requests; every game in a response over the deadline is forfeited (a loss)
- ✓execution budget (projected) · projected from 124 of 3,722 requests at concurrency 1: mean 0.855 s per request x 3,722, as many as one complete 3,000-game grade sent; grading sums every request's in-sandbox time
- ✓healthy after requests
- 3passed2h 50m into the runtook 2m 59s
startup 1.8 s of 300 s; 2/2 gates passed; The speed gates were measured on a sample and the execution budget projected from it: 124 of a complete grade's 3,722 POST /act requests, from the first decision rounds of the public games rather than whole games.
7 gates
- ✓review prerequisites
- ✓startup · GET /health answered 200 after 1.8 s
- ✓POST /act, round 1 · 32 opening decisions answered with legal actions, 1.02 s round trip on this side with the transport: [5, 7, 3, 7, 0, 5, 7, 11, 9, 9, 7, 0, 9, 11, 1, 5, 9, 5, 5, 7, 5, 11, 11, 9, 0, 11, 5, 0, 4, 11, 8, 9]
- ✓battle slice · 124 POST /act batches over 124 decision rounds, 32 games at a time: the first 32 public games (teams from development seed 91827461), 88 finished games replaced by further development-seed matchups, against the pinned Foul Play: 3523 observations, 22-32 per batch, up to 323 KB per request, every answered action legal; not scored
- ✓move response within 1 s · slowest of 124 POST /act responses (22-32 observations, up to 323 KB; the first a cold batch of 32) on the trusted in-sandbox move timer; median 0.852 s, p95 0.857 s. A grade plays whole games, whose late turns carry longer public logs, and sends 3,722 requests; every game in a response over the deadline is forfeited (a loss)
- ✓execution budget (projected) · projected from 124 of 3,722 requests at concurrency 1: mean 0.852 s per request x 3,722, as many as one complete 3,000-game grade sent; grading sums every request's in-sandbox time
- ✓healthy after requests
- 4passed6h 34m into the runtook 3m 2s
startup 1.8 s of 300 s; 2/2 gates passed; The speed gates were measured on a sample and the execution budget projected from it: 124 of a complete grade's 3,722 POST /act requests, from the first decision rounds of the public games rather than whole games.
7 gates
- ✓review prerequisites
- ✓startup · GET /health answered 200 after 1.8 s
- ✓POST /act, round 1 · 32 opening decisions answered with legal actions, 1.01 s round trip on this side with the transport: [1, 7, 1, 7, 0, 0, 7, 11, 10, 9, 7, 0, 9, 11, 1, 5, 9, 11, 0, 7, 5, 11, 11, 9, 0, 11, 5, 3, 4, 11, 8, 10]
- ✓battle slice · 124 POST /act batches over 124 decision rounds, 32 games at a time: the first 32 public games (teams from development seed 91827461), 76 finished games replaced by further development-seed matchups, against the pinned Foul Play: 3559 observations, 23-32 per batch, up to 348 KB per request, every answered action legal; not scored
- ✓move response within 1 s · slowest of 124 POST /act responses (23-32 observations, up to 348 KB; the first a cold batch of 32) on the trusted in-sandbox move timer; median 0.853 s, p95 0.856 s. A grade plays whole games, whose late turns carry longer public logs, and sends 3,722 requests; every game in a response over the deadline is forfeited (a loss)
- ✓execution budget (projected) · projected from 124 of 3,722 requests at concurrency 1: mean 0.853 s per request x 3,722, as many as one complete 3,000-game grade sent; grading sums every request's in-sandbox time
- ✓healthy after requests
The agent's last message before its session ended.
The final H100 self-check passed: all 3,559 sampled decisions were legal, and the slowest batch took 0.858 seconds. I’m submitting the unchanged, checked bundle now.