Submit evaluation results for TimeRSI (part 2)

# Submit evaluation results for TimeRSI

TimeRSI: Time-Series Foundation Model Deployment Needs a Self-Improving Harness

Time-series foundation models such as TimesFM-3.0 offer strong zero-shot forecasts, but deploying them out of the box is often suboptimal because it overlooks dataset semantics and feedback from past forecast errors. We introduce TimeRSI, a self-improving harness that pairs a frozen foundation model with code agents to evolve the forecasting pipeline around it. In each iteration, agents revise features, model inputs, or forecast corrections, run the model, inspect validation errors, and use the results to guide the next revision. At deployment, a selector uses only outcomes from previously completed windows to choose among the resulting strategies for each new forecast. We confirm that all submitted predictions were generated by TimeRSI’s fully automated deployment loop, customized for each forecasting task.

For more details, please see [TimeRSI on GitHub](https://github.com/haoxin1998/TimeRSI).

results/TimeRSI/current_velocity/15T/long/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:933c76652bfefea4f6c729bacb540bf8f441b419b6218882adc524c7c4d5e249
+size 10853

results/TimeRSI/current_velocity/15T/long/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18acee9c22f817232a76aeb460337630916c4cb8700490b740eb69eeaa687849
+size 449340

results/TimeRSI/current_velocity/15T/medium/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/15T/medium",
+  "num_series": 5,
+  "num_windows": 15,
+  "num_variates": 6,
+  "prediction_length": 96,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "15T",
+  "seasonality": 96,
+  "context_length": 17857,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/15T/medium/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3036c9bfa6b7cde564a6c6cf30a6015e930db3eb361a45afa795cca33d79918
+size 22316

results/TimeRSI/current_velocity/15T/medium/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0af813abc0aff6f0612cce31da9a26c361bff83b3971463c100607fa9268ab8
+size 473511

results/TimeRSI/current_velocity/15T/short/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/15T/short",
+  "num_series": 5,
+  "num_windows": 120,
+  "num_variates": 6,
+  "prediction_length": 12,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "15T",
+  "seasonality": 96,
+  "context_length": 17941,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/15T/short/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc5051c2507cfa3c1fc21edda00e07a38e38167823a1c31f6f9b9b8ecda4094b
+size 163602

results/TimeRSI/current_velocity/15T/short/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e71d51b5b3b1a4e2969b97d9f75efee4da2beb8703f69cce6592d73f5919992a
+size 513067

results/TimeRSI/current_velocity/20T/long/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/20T/long",
+  "num_series": 27,
+  "num_windows": 5,
+  "num_variates": 6,
+  "prediction_length": 216,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "20T",
+  "seasonality": 72,
+  "context_length": 16543,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/20T/long/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:868ee832830f91ac319fddc62bddacf91349016455512665a7812ac0c3500bc1
+size 49702

results/TimeRSI/current_velocity/20T/long/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f40132d9cf0652e13427caa621c041017dd6215ed64e7d3c60843053c5ede13a
+size 2031356

results/TimeRSI/current_velocity/20T/medium/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/20T/medium",
+  "num_series": 27,
+  "num_windows": 15,
+  "num_variates": 6,
+  "prediction_length": 72,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "20T",
+  "seasonality": 72,
+  "context_length": 16687,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/20T/medium/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a24a54bf6b3037789749f04827070459adfa15250dee66dce707fed9cc739fe
+size 143643

results/TimeRSI/current_velocity/20T/medium/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09295420a3ba47de090b4f9cf9030282e9eab19279e072c489d04dd90d7e6fed
+size 2094614

results/TimeRSI/current_velocity/20T/short/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/20T/short",
+  "num_series": 27,
+  "num_windows": 120,
+  "num_variates": 6,
+  "prediction_length": 9,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "20T",
+  "seasonality": 72,
+  "context_length": 16750,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/20T/short/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fbd84ce86167e57fbd4f77d642d30c0084345b4443da3b7695df016cfdcc545b
+size 1112928

results/TimeRSI/current_velocity/20T/short/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58e13b4956c78a7d1b29a809e047712904521a9e6d3316bff58af70bcf6154d4
+size 2227659

results/TimeRSI/current_velocity/5T/long/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/5T/long",
+  "num_series": 1,
+  "num_windows": 5,
+  "num_variates": 6,
+  "prediction_length": 864,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "5T",
+  "seasonality": 288,
+  "context_length": 25622,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/5T/long/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3af18a6ecc0afec3db821f52eebc906304e80be2b7c1a1d334e0101a4d29e6a5
+size 3498

results/TimeRSI/current_velocity/5T/long/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bee9caed8ae83293874bc6ddd61b50ca2f699390b5a925ecbc530c1b4d440f19
+size 222995

results/TimeRSI/current_velocity/5T/medium/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/5T/medium",
+  "num_series": 1,
+  "num_windows": 15,
+  "num_variates": 6,
+  "prediction_length": 288,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "5T",
+  "seasonality": 288,
+  "context_length": 26198,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/5T/medium/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a48bb3ccb6da8e0083f86c27fc3e978d7bf5374071e62b634aea57a8bfae7679
+size 7280

results/TimeRSI/current_velocity/5T/medium/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f3fcc84bee3db0426c77fd691b7960b0c85122de536cdc95015fdfbb19cd9bc
+size 231720

results/TimeRSI/current_velocity/5T/short/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/5T/short",
+  "num_series": 1,
+  "num_windows": 120,
+  "num_variates": 6,
+  "prediction_length": 36,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "5T",
+  "seasonality": 288,
+  "context_length": 26450,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/5T/short/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a33482ce3684780f68b59f291c05e9bb94bb2f573c07745cdf0b32c5c106af3
+size 43302

results/TimeRSI/current_velocity/5T/short/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff90f5d493e4f787517bdc5511f355398dae8c52808a0aeb2e0504827899a0f5
+size 254939

results/TimeRSI/current_velocity/H/long/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/H/long",
+  "num_series": 21,
+  "num_windows": 2,
+  "num_variates": 6,
+  "prediction_length": 336,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "H",
+  "seasonality": 24,
+  "context_length": 5251,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/H/long/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54ca19fe73e1d51bb884d2eac23cf96a2e0d7284ea740940200350e73b8f8341
+size 16949

results/TimeRSI/current_velocity/H/long/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd9dba8ca8a686717ac7c2c03b0f0e5842960ecce48989fc84b881ceecfaa005
+size 1080201

results/TimeRSI/current_velocity/H/medium/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/H/medium",
+  "num_series": 21,
+  "num_windows": 4,
+  "num_variates": 6,
+  "prediction_length": 168,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "H",
+  "seasonality": 24,
+  "context_length": 5419,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/H/medium/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e666b441554101216c4de7fefde7d8fb4047147d42647d0b17521813297fdf46
+size 31665

results/TimeRSI/current_velocity/H/medium/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d1659a6c8e252e160199b53c695a211f725747055af0b18ca7085499f78ed62
+size 1103402

results/TimeRSI/current_velocity/H/short/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "current_velocity/H/short",
+  "num_series": 21,
+  "num_windows": 28,
+  "num_variates": 6,
+  "prediction_length": 24,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "H",
+  "seasonality": 24,
+  "context_length": 5563,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/current_velocity/H/short/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c1d25bd747763e7a682479375b4b020fb652f0752dca598c979db10ecbfb3de
+size 205755

results/TimeRSI/current_velocity/H/short/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edfc8e8a4bea058f368302c5157675d4753c8bf23aa9eba88ffc876f592e62e4
+size 1144453

results/TimeRSI/epf_electricity_price/H/long/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "epf_electricity_price/H/long",
+  "num_series": 5,
+  "num_windows": 15,
+  "num_variates": 1,
+  "prediction_length": 168,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "H",
+  "seasonality": 24,
+  "context_length": 52248,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/epf_electricity_price/H/long/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c1ed26bb24e561a4a85346542415b0f77fcfd2c1fe47e2b151469e7309cd6bc
+size 5292

results/TimeRSI/epf_electricity_price/H/long/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b26411e12ba73a5c7abaeb77d6cb61d19ee8a684fec4a9732c442a0a3c6533ea
+size 176431

results/TimeRSI/epf_electricity_price/H/medium/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "epf_electricity_price/H/medium",
+  "num_series": 5,
+  "num_windows": 35,
+  "num_variates": 1,
+  "prediction_length": 72,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "H",
+  "seasonality": 24,
+  "context_length": 52344,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/epf_electricity_price/H/medium/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2fe3bbd80fd5967b300dafe267cd7e5a28ca121b325dbc120a4baa480c49b78
+size 12173

results/TimeRSI/epf_electricity_price/H/medium/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e04c477bfc1313989ab9a5d52204b8ed1f952e16710ac9e9133b1ab8c33fb26
+size 176913

results/TimeRSI/epf_electricity_price/H/short/config.json

@@ -0,0 +1,34 @@
+{
+  "dataset_config": "epf_electricity_price/H/short",
+  "num_series": 5,
+  "num_windows": 105,
+  "num_variates": 1,
+  "prediction_length": 24,
+  "num_quantiles": 9,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "freq": "H",
+  "seasonality": 24,
+  "context_length": 52392,
+  "metric_names": [
+    "MSE",
+    "MAE",
+    "RMSE",
+    "MAPE",
+    "sMAPE",
+    "MASE",
+    "ND",
+    "CRPS"
+  ],
+  "prediction_scale_factor": 1.0,
+  "model_name": "TimeRSI"
+}

results/TimeRSI/epf_electricity_price/H/short/metrics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c5ecea010d235e853428b4ec502bc2dc8c993552764c54009f28e9e1b9c8c17
+size 25393

results/TimeRSI/epf_electricity_price/H/short/predictions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b39db47040f75424094ef710a193866f399cf098312a2880f19a7cfc05e3f05
+size 177449