Publish verified sanitized model artifacts

.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

DATASET.md

@@ -0,0 +1,5 @@
+# Dataset and generator record
+
+This smoke used the deterministic `jump.track-h-dataset/v1` generator at source commit `bb35da1ea280c985f2b438a4e3f97962c08050d7`. It generated two train and two test worlds from root seed 7613, with six 2D objects, hidden non-trivial binary types, exponents 1 and 2, and domain-separated RNG streams for hidden types, appearance, initial state, and law. Each episode had three simulation frames. Train and test world seeds were disjoint.
+
+Dataset SHA-256: `12e1a5622b7d3c948aa349983159b89eb068503783285ae9085a5c9ba0ac6cb0`. The generated rows are reproducible from the manifest and source; raw rows are not duplicated in this model repository.

PRIVACY_AUDIT.json

@@ -0,0 +1 @@
+{"credential_pattern_findings":0,"current_tree_scan":"passed","history_commit_count_expected":1,"local_path_findings":0,"personal_identity_findings":0,"repository":"jump-world-model","schema_version":"jump.hf-privacy-audit/v1","secret_values_recorded":false,"source_archive_repository":"Perfect7613/jump-world-model-private-archive-20260813","source_history_migrated":false,"visibility_target":"public_after_independent_verification"}

README.md

@@ -0,0 +1,60 @@
+---
+license: other
+tags:
+- physics
+- world-model
+- research
+---
+
+# JUMP hidden-relation dynamics smoke
+
+This public artifact contains the single learned scale from a one-step Track H infrastructure smoke. The model multiplies an analytic pairwise-force basis by a learned scalar. The basis uses relative 2D geometry, the known same/different hidden relation, the categorical force law, and its exponent. This is not a Gemma adapter, a general world model, or evidence for a mechanistic claim.
+
+## Result
+
+One H100 optimization step on two training worlds produced held-out force NRMSE 0.9444 on two test worlds, compared with 1.0 for a zero-force persistence baseline (5.56% relative improvement). The run is below the preregistered G2 direction of at least 20% improvement across three seeds. It passed only the one-step execution and artifact gates.
+
+The runner recorded 7.49344 seconds wall duration and 0.230006 seconds inside CUDA training. The pinned-rate ledger estimate was USD 0.0082203; this is not a Modal invoice. Peak PyTorch allocated memory was approximately 7 KiB because the checkpoint has one trainable scalar.
+
+## Files
+
+- `model.safetensors`: JUMP-produced weight only; SHA-256 `6702724b6dc658764146a270a45e4eaa9110f92303757069a9b0ef0187d6e967`.
+- `config.json`: architecture and feature contract.
+- `training-manifest.json`, `run-result.json`, `phase-result.json`: immutable execution inputs and results.
+- `runner-hashes.sha256`, `SHA256SUMS`: runner and publication checksums.
+- `DATASET.md`, `training.json`, `environment.json`, `requirements.lock`: generator, metrics, and runtime record.
+- `references.json`: source revision and the pinned Gemma target that was **not** used to train this checkpoint. No tokenizer files or Gemma weights are included.
+
+Optimizer state, caches, credentials, and raw execution logs are intentionally excluded.
+
+## Intended use
+
+Use this artifact to verify loading, provenance, and the deterministic Track H training path. It may also initialize reproduction of the exact one-step smoke. Do not use it as a scientific result, a force-law discovery model, or a component for real-world scientific, medical, or safety decisions.
+
+## Reproduction
+
+```bash
+git clone https://github.com/perfect7613/jump-research.git
+cd jump-research
+git checkout bb35da1ea280c985f2b438a4e3f97962c08050d7
+python -m pip install -e '.[test]'
+pytest -q
+# Inspect the fixed H100 entrypoint and manifest before any paid run.
+modal deploy -m jump_runner.modal_app::app
+```
+
+A paid rerun requires the repository safety checks and explicit authorization. The checked-in CPU and T4 manifests can be validated without launching a full matrix.
+
+## Limitations
+
+The checkpoint has one scalar parameter, one optimization step, one root seed, four worlds, and no uncertainty interval. The analytic input basis supplies the hidden relation and force-law structure, so the result does not show that a model inferred either structure. The run does not test A/B/C/C-prime language-model behavior, latent injection, OOD transfer, replication, or mediation.
+
+## License and attribution
+
+The source repository currently has no selected license; its README states that all rights are reserved pending licensing decisions. This model repository therefore grants no additional reuse license. The artifact contains no Gemma weights or tokenizer files. `references.json` records the separately licensed Gemma checkpoint considered for later adapter work. Review the source repository and any future base-model terms before reuse or redistribution.
+
+## Three-seed pilot update
+
+A subsequent fixed H100 pilot ran 120 optimization steps independently at seeds 7613, 7614, and 7615, using eight train and four held-out test worlds per seed. Held-out force NRMSE values were 0.0000994703, 0.0000994727, and 0.0000994612. The mean was 0.0000994680 with sample standard deviation 0.0000000060; all three correspond to about 99.9901% improvement over the zero-force baseline. With only three deterministic seeds, this dispersion is descriptive and is not a confidence interval.
+
+The >=20% NRMSE-improvement direction passed for all three seeds. Full G2 did **not** pass because this pilot did not measure hidden-relation probe AUC or its confidence bound, and the analytic input basis supplies the hidden relation and force-law structure. The result remains Track H engineering evidence only. `pilot/verification.json` binds every model, dataset, and checkpoint hash after checking the runner checksum manifests. Optimizer checkpoints remain on the private Modal Volume and are not uploaded.

SHA256SUMS

@@ -0,0 +1,34 @@
+9985408d4457ca1828ce92d83ac85ada620ed13694ce1477fbf83832fb8ab219  DATASET.md
+94fe9c56c90c5545e20cfcb34bf6daf760852402cd4879379f92bf89d52a7413  PRIVACY_AUDIT.json
+b6f0ccfc6f6344fac6783b1e9690b1362d9cd30477dcc042a6bae77813c9d98c  README.md
+01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10  config.json
+07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc  environment.json
+6702724b6dc658764146a270a45e4eaa9110f92303757069a9b0ef0187d6e967  model.safetensors
+e5e054bbacf651742e487f86cd95ea70ecdc4a42792341f7388152e6aa68b9b4  phase-result.json
+26fb5a3ccaae45ff3e4e9577b37abb3e8b03aadccc004bc67fb00a178f843fed  pilot/phase-result.json
+01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10  pilot/seed-7613/config.json
+07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc  pilot/seed-7613/environment.json
+a14a47f6baeb07da486d18e794df7fe3d0978f7b377644a2020e0dd15b88f4be  pilot/seed-7613/model.safetensors
+b8862d2478cd7031cd1a149812b269edd6c688483dccc7095a740f47bab89d9d  pilot/seed-7613/run-result.json
+c11b0c5a9fed6e09e67faf1805f0d928d3434019b3a2d3ce7dfe847af9a794c6  pilot/seed-7613/runner-hashes.sha256
+93236fe32120997200feea335490e118481a7961ff0c3066352acf821eab1225  pilot/seed-7613/training.json
+01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10  pilot/seed-7614/config.json
+07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc  pilot/seed-7614/environment.json
+dfb965ca71ef4ecb55d48c7a5fe571bfdca7d2ba8c70bba95f04b37ff1063493  pilot/seed-7614/model.safetensors
+6b5cbbdcb098e2ea7006cd7e2810b5e1de0b5fc41b9858ca26b6d508e6f6b449  pilot/seed-7614/run-result.json
+2f9358d37dbd7a47a58c2e3c98f50ae9045eba169af3a634bad8d061e81da2e8  pilot/seed-7614/runner-hashes.sha256
+9a3727e676730834421f6df48eb48818f5e7eb58457267b8458271c3bf033d69  pilot/seed-7614/training.json
+01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10  pilot/seed-7615/config.json
+07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc  pilot/seed-7615/environment.json
+733526058f6ed7cf9af980d7d955543fca0c7ccee80e2f737631d2d2fe285843  pilot/seed-7615/model.safetensors
+d3a939b654500d9c072b98231f86b008028155709d18acfb6834c91350968e6f  pilot/seed-7615/run-result.json
+4160198369d5733883941a3eccfba3b9da270ce1e2f8fa4d2a7af37081ce766c  pilot/seed-7615/runner-hashes.sha256
+5626ce7c993a6a11ad5f2c46a4ada66579f883fe4a8709f545785a5e75d02ebf  pilot/seed-7615/training.json
+e0ec6fe97c906cee403e255dcf63fe56d3521a4fabd3b910f61d04ee8d8d40cf  pilot/training-manifest.json
+3d4505f199920b8f08170cabf2480bf0fadf4cabd0f1cdf7c5b29bf443330d15  pilot/verification.json
+647802b7bec2a9c5c99ffcb462316b577db0decf681416a486fa605a2374e109  references.json
+89ce6efb9548d1c8565d301ae87883f48e30baa6566679cbd08298de9a0ddfc2  requirements.lock
+32b83eba725742535fd803428a48bc219d28ad6f83909fce74937a8d5be38c78  run-result.json
+617c79e27d15afbb76449a43ee1576b99886a3a6198ef2ad411d0ef0c6cee213  runner-hashes.sha256
+de834d1cb0f2b48e67ab294c21dca7f8c5350dca6d36e2cd7f002ada5a81ac56  training-manifest.json
+67d777ab93cd35a957a6f6ee014c057ed0f5b24ccc24b7930328d9570a1ca088  training.json

config.json

@@ -0,0 +1 @@
+{"architecture":"learned_scalar_times_hidden_relation_force_basis","base_model":null,"generator":"jump.track-h-dataset/v1","input_features":["unit_coefficient_force_basis_x","unit_coefficient_force_basis_y"],"output_features":["force_x","force_y"],"parameter_names":["scale"],"schema_version":"jump.track-h-world-model-config/v1","tokenizer":null}

environment.json

@@ -0,0 +1 @@
+{"deterministic_algorithms":true,"device_type":"cuda","python_torch":"2.11.0","safetensors":"0.6.2"}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6702724b6dc658764146a270a45e4eaa9110f92303757069a9b0ef0187d6e967
+size 260

phase-result.json

@@ -0,0 +1 @@
+{"completed_at":"2026-08-12T22:02:36.918867+00:00","estimated_cost_usd":0.008220303679999999,"forecast_cost_usd":0.32909999999999995,"gpu":"H100","mechanistic_escalation_evidence":false,"phase_id":"authorized-h100-training-smoke","runtime_seconds":7.49344,"schema_version":"jump.phase-result/v1","status":"passed"}

pilot/phase-result.json

@@ -0,0 +1 @@
+{"completed_at":"2026-08-12T22:07:28.624443+00:00","estimated_cost_usd":0.046498915271,"forecast_cost_usd":1.1847599999999998,"full_g2_passed":false,"g2_improvement_direction_passed":true,"mechanistic_evidence":false,"phase_id":"authorized-h100-three-seed-pilot","relative_improvements":[0.9999005297431722,0.9999005273493822,0.9999005387871875],"runtime_seconds":42.387343,"schema_version":"jump.phase-result/v1","seeds":[7613,7614,7615],"status":"passed"}

pilot/seed-7613/config.json

@@ -0,0 +1 @@
+{"architecture":"learned_scalar_times_hidden_relation_force_basis","base_model":null,"generator":"jump.track-h-dataset/v1","input_features":["unit_coefficient_force_basis_x","unit_coefficient_force_basis_y"],"output_features":["force_x","force_y"],"parameter_names":["scale"],"schema_version":"jump.track-h-world-model-config/v1","tokenizer":null}

pilot/seed-7613/environment.json

@@ -0,0 +1 @@
+{"deterministic_algorithms":true,"device_type":"cuda","python_torch":"2.11.0","safetensors":"0.6.2"}

pilot/seed-7613/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a14a47f6baeb07da486d18e794df7fe3d0978f7b377644a2020e0dd15b88f4be
+size 260

pilot/seed-7613/run-result.json

@@ -0,0 +1 @@
+{"artifacts":[{"media_type":"application/octet-stream","name":"environment.json","path":"attempts/0001/artifacts/environment.json","sha256":"07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc"},{"media_type":"application/octet-stream","name":"model-config.json","path":"attempts/0001/artifacts/model-config.json","sha256":"01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10"},{"media_type":"application/octet-stream","name":"model.safetensors","path":"attempts/0001/artifacts/model.safetensors","sha256":"a14a47f6baeb07da486d18e794df7fe3d0978f7b377644a2020e0dd15b88f4be"},{"media_type":"application/octet-stream","name":"training.json","path":"attempts/0001/artifacts/training.json","sha256":"93236fe32120997200feea335490e118481a7961ff0c3066352acf821eab1225"}],"attempt":1,"duration_seconds":19.084096,"error":null,"exit_code":0,"finished_at":"2026-08-12T22:06:59.498825+00:00","metrics":[{"name":"world_model.train.initial_mse","split":"train","value":0.045661699026823044},{"name":"world_model.train.final_mse","split":"train","value":7.301780513557787e-09},{"name":"world_model.test.force_nrmse","split":"test","value":9.947025682777166e-05},{"name":"world_model.test.persistence_nrmse","split":"test","value":1.0},{"name":"world_model.test.relative_improvement","split":"test","value":0.9999005297431722},{"name":"world_model.training_seconds","split":"all","value":1.6239877559999982},{"name":"world_model.training_steps","split":"all","value":120},{"name":"world_model.peak_memory_gb","split":"all","value":1.1920928955078125e-05},{"name":"world_model.train.sample_count","split":"train","value":192},{"name":"world_model.test.sample_count","split":"test","value":96}],"provenance":{"code_version":"90b41dbaaa8fae2813cea280ada6d62d37a7a06b","manifest_sha256":"e0ec6fe97c906cee403e255dcf63fe56d3521a4fabd3b910f61d04ee8d8d40cf","run_id":"hidden-relation-dynamics-seed-7613"},"retryable":false,"schema_version":"jump.run-result/v1","started_at":"2026-08-12T22:06:40.323754+00:00","status":"completed","track_h":{"dataset_sha256":"cdca04d5026e2b240cd9a41d3a30656c0d44e8c98c3c827f0f6f8b7b5eee75e5","mechanistic_evidence":false,"task":"jump.track-h-world-model-smoke/v1"}}

pilot/seed-7613/runner-hashes.sha256

@@ -0,0 +1,10 @@
+07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc  artifacts/environment.json
+01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10  artifacts/model-config.json
+a14a47f6baeb07da486d18e794df7fe3d0978f7b377644a2020e0dd15b88f4be  artifacts/model.safetensors
+93236fe32120997200feea335490e118481a7961ff0c3066352acf821eab1225  artifacts/training.json
+22161bee63b1268812cb07eed9fb6cf3284a565cfa88dea3183fd902d822bf22  checkpoint/world-model-checkpoint.pt
+b8862d2478cd7031cd1a149812b269edd6c688483dccc7095a740f47bab89d9d  result.json
+08688ba3e232d43ceec8908c29eb9f9dcb0f179ee5853179c89de5a56541fe29  started.json
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  stderr.log
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  stdout.log
+c7e427d62952526b47460a0af480a4d03cba4aca3b44f0ed1cf65ca553e56a68  task-parameters.json

pilot/seed-7613/training.json

@@ -0,0 +1 @@
+{"config":{"epochs":120,"exponents":[1,2],"learning_rate":0.01,"require_cuda":true,"seed":7613,"simulator":{"bounds":3.0,"coefficient":0.18,"dt":0.04,"softening":0.2,"steps":4},"task":"jump.track-h-world-model-smoke/v1","test_episodes":4,"torch_version":"2.11.0","train_episodes":8},"dataset_sha256":"cdca04d5026e2b240cd9a41d3a30656c0d44e8c98c3c827f0f6f8b7b5eee75e5","device_type":"cuda","final_train_mse":7.301780513557787e-09,"initial_train_mse":0.045661699026823044,"mechanistic_evidence":false,"peak_memory_gb":1.1920928955078125e-05,"persistence_force_nrmse":1.0,"relative_improvement":0.9999005297431722,"schema_version":"jump.track-h-world-model-smoke/v1","test_force_nrmse":9.947025682777166e-05,"test_sample_count":96,"torch_version":"2.11.0","train_sample_count":192,"training_seconds":1.6239877559999982}

pilot/seed-7614/config.json

@@ -0,0 +1 @@
+{"architecture":"learned_scalar_times_hidden_relation_force_basis","base_model":null,"generator":"jump.track-h-dataset/v1","input_features":["unit_coefficient_force_basis_x","unit_coefficient_force_basis_y"],"output_features":["force_x","force_y"],"parameter_names":["scale"],"schema_version":"jump.track-h-world-model-config/v1","tokenizer":null}

pilot/seed-7614/environment.json

@@ -0,0 +1 @@
+{"deterministic_algorithms":true,"device_type":"cuda","python_torch":"2.11.0","safetensors":"0.6.2"}

pilot/seed-7614/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfb965ca71ef4ecb55d48c7a5fe571bfdca7d2ba8c70bba95f04b37ff1063493
+size 260

pilot/seed-7614/run-result.json

@@ -0,0 +1 @@
+{"artifacts":[{"media_type":"application/octet-stream","name":"environment.json","path":"attempts/0001/artifacts/environment.json","sha256":"07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc"},{"media_type":"application/octet-stream","name":"model-config.json","path":"attempts/0001/artifacts/model-config.json","sha256":"01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10"},{"media_type":"application/octet-stream","name":"model.safetensors","path":"attempts/0001/artifacts/model.safetensors","sha256":"dfb965ca71ef4ecb55d48c7a5fe571bfdca7d2ba8c70bba95f04b37ff1063493"},{"media_type":"application/octet-stream","name":"training.json","path":"attempts/0001/artifacts/training.json","sha256":"9a3727e676730834421f6df48eb48818f5e7eb58457267b8458271c3bf033d69"}],"attempt":1,"duration_seconds":11.481514,"error":null,"exit_code":0,"finished_at":"2026-08-12T22:07:13.585874+00:00","metrics":[{"name":"world_model.train.initial_mse","split":"train","value":0.025443460792303085},{"name":"world_model.train.final_mse","split":"train","value":4.06866806912376e-09},{"name":"world_model.test.force_nrmse","split":"test","value":9.947265061782673e-05},{"name":"world_model.test.persistence_nrmse","split":"test","value":1.0},{"name":"world_model.test.relative_improvement","split":"test","value":0.9999005273493822},{"name":"world_model.training_seconds","split":"all","value":1.4656814070000053},{"name":"world_model.training_steps","split":"all","value":120},{"name":"world_model.peak_memory_gb","split":"all","value":1.1920928955078125e-05},{"name":"world_model.train.sample_count","split":"train","value":192},{"name":"world_model.test.sample_count","split":"test","value":96}],"provenance":{"code_version":"90b41dbaaa8fae2813cea280ada6d62d37a7a06b","manifest_sha256":"e0ec6fe97c906cee403e255dcf63fe56d3521a4fabd3b910f61d04ee8d8d40cf","run_id":"hidden-relation-dynamics-seed-7614"},"retryable":false,"schema_version":"jump.run-result/v1","started_at":"2026-08-12T22:07:02.015675+00:00","status":"completed","track_h":{"dataset_sha256":"aba36d3690cc72f74834519ba066cb267254ad220f608bd3a61d73a3c379ad12","mechanistic_evidence":false,"task":"jump.track-h-world-model-smoke/v1"}}

pilot/seed-7614/runner-hashes.sha256

@@ -0,0 +1,10 @@
+07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc  artifacts/environment.json
+01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10  artifacts/model-config.json
+dfb965ca71ef4ecb55d48c7a5fe571bfdca7d2ba8c70bba95f04b37ff1063493  artifacts/model.safetensors
+9a3727e676730834421f6df48eb48818f5e7eb58457267b8458271c3bf033d69  artifacts/training.json
+7e07cffc1aff9e8d62be4b0229483371f33877acbccac916b5ac999d4906b8df  checkpoint/world-model-checkpoint.pt
+6b5cbbdcb098e2ea7006cd7e2810b5e1de0b5fc41b9858ca26b6d508e6f6b449  result.json
+3c4f1ff82e655dfe6e9c7d33568ec4fddba0b70d388f491d512052ab76f9e7df  started.json
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  stderr.log
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  stdout.log
+21ebb062dd48d9b1e4f55ad6f665efa7af8fc058db1d6dcc37b5bca71a6446ee  task-parameters.json

pilot/seed-7614/training.json

@@ -0,0 +1 @@
+{"config":{"epochs":120,"exponents":[1,2],"learning_rate":0.01,"require_cuda":true,"seed":7614,"simulator":{"bounds":3.0,"coefficient":0.18,"dt":0.04,"softening":0.2,"steps":4},"task":"jump.track-h-world-model-smoke/v1","test_episodes":4,"torch_version":"2.11.0","train_episodes":8},"dataset_sha256":"aba36d3690cc72f74834519ba066cb267254ad220f608bd3a61d73a3c379ad12","device_type":"cuda","final_train_mse":4.06866806912376e-09,"initial_train_mse":0.025443460792303085,"mechanistic_evidence":false,"peak_memory_gb":1.1920928955078125e-05,"persistence_force_nrmse":1.0,"relative_improvement":0.9999005273493822,"schema_version":"jump.track-h-world-model-smoke/v1","test_force_nrmse":9.947265061782673e-05,"test_sample_count":96,"torch_version":"2.11.0","train_sample_count":192,"training_seconds":1.4656814070000053}

pilot/seed-7615/config.json

@@ -0,0 +1 @@
+{"architecture":"learned_scalar_times_hidden_relation_force_basis","base_model":null,"generator":"jump.track-h-dataset/v1","input_features":["unit_coefficient_force_basis_x","unit_coefficient_force_basis_y"],"output_features":["force_x","force_y"],"parameter_names":["scale"],"schema_version":"jump.track-h-world-model-config/v1","tokenizer":null}

pilot/seed-7615/environment.json

@@ -0,0 +1 @@
+{"deterministic_algorithms":true,"device_type":"cuda","python_torch":"2.11.0","safetensors":"0.6.2"}

pilot/seed-7615/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:733526058f6ed7cf9af980d7d955543fca0c7ccee80e2f737631d2d2fe285843
+size 260

pilot/seed-7615/run-result.json

@@ -0,0 +1 @@
+{"artifacts":[{"media_type":"application/octet-stream","name":"environment.json","path":"attempts/0001/artifacts/environment.json","sha256":"07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc"},{"media_type":"application/octet-stream","name":"model-config.json","path":"attempts/0001/artifacts/model-config.json","sha256":"01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10"},{"media_type":"application/octet-stream","name":"model.safetensors","path":"attempts/0001/artifacts/model.safetensors","sha256":"733526058f6ed7cf9af980d7d955543fca0c7ccee80e2f737631d2d2fe285843"},{"media_type":"application/octet-stream","name":"training.json","path":"attempts/0001/artifacts/training.json","sha256":"5626ce7c993a6a11ad5f2c46a4ada66579f883fe4a8709f545785a5e75d02ebf"}],"attempt":1,"duration_seconds":11.821733,"error":null,"exit_code":0,"finished_at":"2026-08-12T22:07:26.974330+00:00","metrics":[{"name":"world_model.train.initial_mse","split":"train","value":0.029812103137373924},{"name":"world_model.train.final_mse","split":"train","value":4.767301220454101e-09},{"name":"world_model.test.force_nrmse","split":"test","value":9.946121281245723e-05},{"name":"world_model.test.persistence_nrmse","split":"test","value":1.0},{"name":"world_model.test.relative_improvement","split":"test","value":0.9999005387871875},{"name":"world_model.training_seconds","split":"all","value":1.5545569179999958},{"name":"world_model.training_steps","split":"all","value":120},{"name":"world_model.peak_memory_gb","split":"all","value":1.1920928955078125e-05},{"name":"world_model.train.sample_count","split":"train","value":192},{"name":"world_model.test.sample_count","split":"test","value":96}],"provenance":{"code_version":"90b41dbaaa8fae2813cea280ada6d62d37a7a06b","manifest_sha256":"e0ec6fe97c906cee403e255dcf63fe56d3521a4fabd3b910f61d04ee8d8d40cf","run_id":"hidden-relation-dynamics-seed-7615"},"retryable":false,"schema_version":"jump.run-result/v1","started_at":"2026-08-12T22:07:15.074451+00:00","status":"completed","track_h":{"dataset_sha256":"cd804a9e2a549e700d46b0e739fed51e97c2ba793e078c04ee8f6df3542b522d","mechanistic_evidence":false,"task":"jump.track-h-world-model-smoke/v1"}}

pilot/seed-7615/runner-hashes.sha256

@@ -0,0 +1,10 @@
+07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc  artifacts/environment.json
+01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10  artifacts/model-config.json
+733526058f6ed7cf9af980d7d955543fca0c7ccee80e2f737631d2d2fe285843  artifacts/model.safetensors
+5626ce7c993a6a11ad5f2c46a4ada66579f883fe4a8709f545785a5e75d02ebf  artifacts/training.json
+9690f7a2ceb556626467c0c6f9051353a66c9ae5c4eea698aa5a9c2fdad1db3c  checkpoint/world-model-checkpoint.pt
+d3a939b654500d9c072b98231f86b008028155709d18acfb6834c91350968e6f  result.json
+0f0626475f685ff0af1fd8d72ff16a5d0c29ff48254902b32070d831be179031  started.json
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  stderr.log
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  stdout.log
+b3965b4d126b520f4d553f599198133138cb71de84991663b65f934638d78fe8  task-parameters.json

pilot/seed-7615/training.json

@@ -0,0 +1 @@
+{"config":{"epochs":120,"exponents":[1,2],"learning_rate":0.01,"require_cuda":true,"seed":7615,"simulator":{"bounds":3.0,"coefficient":0.18,"dt":0.04,"softening":0.2,"steps":4},"task":"jump.track-h-world-model-smoke/v1","test_episodes":4,"torch_version":"2.11.0","train_episodes":8},"dataset_sha256":"cd804a9e2a549e700d46b0e739fed51e97c2ba793e078c04ee8f6df3542b522d","device_type":"cuda","final_train_mse":4.767301220454101e-09,"initial_train_mse":0.029812103137373924,"mechanistic_evidence":false,"peak_memory_gb":1.1920928955078125e-05,"persistence_force_nrmse":1.0,"relative_improvement":0.9999005387871875,"schema_version":"jump.track-h-world-model-smoke/v1","test_force_nrmse":9.946121281245723e-05,"test_sample_count":96,"torch_version":"2.11.0","train_sample_count":192,"training_seconds":1.5545569179999958}

pilot/training-manifest.json

@@ -0,0 +1 @@
+{"description":"Three-seed Track H dynamics pilot for the G2 improvement direction only; not a full matrix or mechanistic evidence.","experiment_id":"track-h-world-model-h100-three-seed-pilot","launch_policy":{"allow_full_matrix":false,"allow_h100":true,"narrow_training_pilot_only":true},"phases":[{"budget":{"allowed_gpu_types":["H100"],"gpu_hourly_cost_usd":{"H100":3.9492},"max_concurrent_gpus":1,"max_cost_usd":1.1847599999999998,"max_runtime_seconds":600},"id":"authorized-h100-three-seed-pilot","runs":[{"id":"hidden-relation-dynamics-seed-7613","resources":{"gpu":"H100","timeout_seconds":180},"retry":{"max_attempts":2},"task":{"module":"jump_benchmark.train","parameters":{"epochs":120,"exponents":[1,2],"learning_rate":0.01,"require_cuda":true,"seed":7613,"simulator":{"bounds":3.0,"coefficient":0.18,"dt":0.04,"softening":0.2,"steps":4},"task":"jump.track-h-world-model-smoke/v1","test_episodes":4,"torch_version":"2.11.0","train_episodes":8}}},{"id":"hidden-relation-dynamics-seed-7614","resources":{"gpu":"H100","timeout_seconds":180},"retry":{"max_attempts":2},"task":{"module":"jump_benchmark.train","parameters":{"epochs":120,"exponents":[1,2],"learning_rate":0.01,"require_cuda":true,"seed":7614,"simulator":{"bounds":3.0,"coefficient":0.18,"dt":0.04,"softening":0.2,"steps":4},"task":"jump.track-h-world-model-smoke/v1","test_episodes":4,"torch_version":"2.11.0","train_episodes":8}}},{"id":"hidden-relation-dynamics-seed-7615","resources":{"gpu":"H100","timeout_seconds":180},"retry":{"max_attempts":2},"task":{"module":"jump_benchmark.train","parameters":{"epochs":120,"exponents":[1,2],"learning_rate":0.01,"require_cuda":true,"seed":7615,"simulator":{"bounds":3.0,"coefficient":0.18,"dt":0.04,"softening":0.2,"steps":4},"task":"jump.track-h-world-model-smoke/v1","test_episodes":4,"torch_version":"2.11.0","train_episodes":8}}}]}],"preregistration":{"layer_allowlist":["not-applicable"],"timepoint_allowlist":["not-applicable"]},"schema_version":"jump.experiments/v1"}

pilot/verification.json

@@ -0,0 +1,41 @@
+{
+  "improvement_mean": 0.999900531959914,
+  "improvement_sample_stdev": 6.032520693511383e-09,
+  "nrmse_max": 9.947265061782673e-05,
+  "nrmse_mean": 9.946804008601855e-05,
+  "nrmse_min": 9.946121281245723e-05,
+  "nrmse_sample_stdev": 6.032520693511383e-09,
+  "per_seed": [
+    {
+      "checkpoint_sha256": "22161bee63b1268812cb07eed9fb6cf3284a565cfa88dea3183fd902d822bf22",
+      "dataset_sha256": "cdca04d5026e2b240cd9a41d3a30656c0d44e8c98c3c827f0f6f8b7b5eee75e5",
+      "duration": 19.084096,
+      "improvement": 0.9999005297431722,
+      "model_sha256": "a14a47f6baeb07da486d18e794df7fe3d0978f7b377644a2020e0dd15b88f4be",
+      "nrmse": 9.947025682777166e-05,
+      "seed": 7613,
+      "training_seconds": 1.6239877559999982
+    },
+    {
+      "checkpoint_sha256": "7e07cffc1aff9e8d62be4b0229483371f33877acbccac916b5ac999d4906b8df",
+      "dataset_sha256": "aba36d3690cc72f74834519ba066cb267254ad220f608bd3a61d73a3c379ad12",
+      "duration": 11.481514,
+      "improvement": 0.9999005273493822,
+      "model_sha256": "dfb965ca71ef4ecb55d48c7a5fe571bfdca7d2ba8c70bba95f04b37ff1063493",
+      "nrmse": 9.947265061782673e-05,
+      "seed": 7614,
+      "training_seconds": 1.4656814070000053
+    },
+    {
+      "checkpoint_sha256": "9690f7a2ceb556626467c0c6f9051353a66c9ae5c4eea698aa5a9c2fdad1db3c",
+      "dataset_sha256": "cd804a9e2a549e700d46b0e739fed51e97c2ba793e078c04ee8f6df3542b522d",
+      "duration": 11.821733,
+      "improvement": 0.9999005387871875,
+      "model_sha256": "733526058f6ed7cf9af980d7d955543fca0c7ccee80e2f737631d2d2fe285843",
+      "nrmse": 9.946121281245723e-05,
+      "seed": 7615,
+      "training_seconds": 1.5545569179999958
+    }
+  ],
+  "verified": true
+}

references.json

@@ -0,0 +1,15 @@
+{
+  "gemma_adapter_target": {
+    "license_metadata": "apache-2.0",
+    "model_id": "google/gemma-4-12B-it",
+    "revision": "707f0a3b8a3c7ad586ed01e27eafbad8a27dd0f7",
+    "used_in_this_checkpoint": false
+  },
+  "schema_version": "jump.model-references/v1",
+  "source_code": {
+    "commit": "bb35da1ea280c985f2b438a4e3f97962c08050d7",
+    "repository": "https://github.com/perfect7613/jump-research"
+  },
+  "tokenizer": null,
+  "trained_base_model": null
+}

requirements.lock

@@ -0,0 +1,4 @@
+huggingface_hub==1.14.0
+numpy==2.3.2
+safetensors==0.6.2
+torch==2.11.0

run-result.json

@@ -0,0 +1 @@
+{"artifacts":[{"media_type":"application/octet-stream","name":"environment.json","path":"attempts/0001/artifacts/environment.json","sha256":"07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc"},{"media_type":"application/octet-stream","name":"model-config.json","path":"attempts/0001/artifacts/model-config.json","sha256":"01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10"},{"media_type":"application/octet-stream","name":"model.safetensors","path":"attempts/0001/artifacts/model.safetensors","sha256":"6702724b6dc658764146a270a45e4eaa9110f92303757069a9b0ef0187d6e967"},{"media_type":"application/octet-stream","name":"training.json","path":"attempts/0001/artifacts/training.json","sha256":"67d777ab93cd35a957a6f6ee014c057ed0f5b24ccc24b7930328d9570a1ca088"}],"attempt":1,"duration_seconds":7.49344,"error":null,"exit_code":0,"finished_at":"2026-08-12T22:02:36.762635+00:00","metrics":[{"name":"world_model.train.initial_mse","split":"train","value":0.039398957043886185},{"name":"world_model.train.final_mse","split":"train","value":0.039398957043886185},{"name":"world_model.test.force_nrmse","split":"test","value":0.9444443583488464},{"name":"world_model.test.persistence_nrmse","split":"test","value":1.0},{"name":"world_model.test.relative_improvement","split":"test","value":0.055555641651153564},{"name":"world_model.training_seconds","split":"all","value":0.23000600100000046},{"name":"world_model.training_steps","split":"all","value":1},{"name":"world_model.peak_memory_gb","split":"all","value":6.67572021484375e-06},{"name":"world_model.train.sample_count","split":"train","value":36},{"name":"world_model.test.sample_count","split":"test","value":36}],"provenance":{"code_version":"bb35da1ea280c985f2b438a4e3f97962c08050d7","manifest_sha256":"de834d1cb0f2b48e67ab294c21dca7f8c5350dca6d36e2cd7f002ada5a81ac56","run_id":"one-step-hidden-relation-dynamics"},"retryable":false,"schema_version":"jump.run-result/v1","started_at":"2026-08-12T22:02:29.228186+00:00","status":"completed","track_h":{"dataset_sha256":"12e1a5622b7d3c948aa349983159b89eb068503783285ae9085a5c9ba0ac6cb0","mechanistic_evidence":false,"task":"jump.track-h-world-model-smoke/v1"}}

runner-hashes.sha256

@@ -0,0 +1,10 @@
+07229f0066680c3a35ac11435dbcd4773505e40b2e66f99e382fbc18371322dc  artifacts/environment.json
+01a148fb4a0a6f193e2dc0bbebeb9af5f974e97077188a91f37a87023dba5e10  artifacts/model-config.json
+6702724b6dc658764146a270a45e4eaa9110f92303757069a9b0ef0187d6e967  artifacts/model.safetensors
+67d777ab93cd35a957a6f6ee014c057ed0f5b24ccc24b7930328d9570a1ca088  artifacts/training.json
+5f4bdda871fb311a82e39369343574c2d314513008c58cd6ce7d3e6befe32c95  checkpoint/world-model-checkpoint.pt
+32b83eba725742535fd803428a48bc219d28ad6f83909fce74937a8d5be38c78  result.json
+5ee84fedc26b23256c6df82f1a0c5dd176c772343a1971a5f688a561fffc627a  started.json
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  stderr.log
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  stdout.log
+de0fff5a80aea0e184b969b8c2196c5d8c38984cc3ea624dc399f22e39a7e427  task-parameters.json

training-manifest.json

@@ -0,0 +1 @@
+{"description":"Human-authorized one-step H100 training smoke; not mechanistic escalation evidence.","experiment_id":"track-h-world-model-h100-one-step","launch_policy":{"allow_full_matrix":false,"allow_h100":true,"narrow_training_smoke_only":true},"phases":[{"budget":{"allowed_gpu_types":["H100"],"gpu_hourly_cost_usd":{"H100":3.9492},"max_concurrent_gpus":1,"max_cost_usd":0.32909999999999995,"max_runtime_seconds":300},"gates":[{"aggregation":"min","id":"exactly-one-training-step","metric":"world_model.training_steps","operator":"==","threshold":1}],"id":"authorized-h100-training-smoke","runs":[{"id":"one-step-hidden-relation-dynamics","resources":{"gpu":"H100","timeout_seconds":300},"retry":{"max_attempts":1},"task":{"module":"jump_benchmark.train","parameters":{"epochs":1,"exponents":[1,2],"learning_rate":0.01,"require_cuda":true,"seed":7613,"simulator":{"bounds":3.0,"coefficient":0.18,"dt":0.04,"softening":0.2,"steps":3},"task":"jump.track-h-world-model-smoke/v1","test_episodes":2,"torch_version":"2.11.0","train_episodes":2}}}]}],"preregistration":{"layer_allowlist":["not-applicable"],"timepoint_allowlist":["not-applicable"]},"schema_version":"jump.experiments/v1"}

training.json

@@ -0,0 +1 @@
+{"config":{"epochs":1,"exponents":[1,2],"learning_rate":0.01,"require_cuda":true,"seed":7613,"simulator":{"bounds":3.0,"coefficient":0.18,"dt":0.04,"softening":0.2,"steps":3},"task":"jump.track-h-world-model-smoke/v1","test_episodes":2,"torch_version":"2.11.0","train_episodes":2},"dataset_sha256":"12e1a5622b7d3c948aa349983159b89eb068503783285ae9085a5c9ba0ac6cb0","device_type":"cuda","final_train_mse":0.039398957043886185,"initial_train_mse":0.039398957043886185,"mechanistic_evidence":false,"peak_memory_gb":6.67572021484375e-06,"persistence_force_nrmse":1.0,"relative_improvement":0.055555641651153564,"schema_version":"jump.track-h-world-model-smoke/v1","test_force_nrmse":0.9444443583488464,"test_sample_count":36,"torch_version":"2.11.0","train_sample_count":36,"training_seconds":0.23000600100000046}