Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -28,21 +28,32 @@ published GIFT scores. Generated by the v10 experiment notebook.
|
|
| 28 |
## Results
|
| 29 |
```
|
| 30 |
=== transfer curve: GM-CRPS by checkpoint (dev-14, mean over seeds) ===
|
| 31 |
-
gm_avg2 gm_avg2_std gm@7500 gm@15000 gm@22500 gm@final gm_final_std long_season params_m
|
| 32 |
-
arm
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
--- tripwire checks ---
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
N0_raw_nat vs N2_vqce_multi_nat: final gap=-0.0136 -> UNRESOLVED (< 2*sigma)
|
| 45 |
N2_vqce_multi_nat vs N4_rawce_nat: final gap=0.0513 -> RESOLVED
|
|
|
|
| 46 |
N0_raw_nat: TRANSFER INVERSION — best at ckpt #2 (0.1938) vs final (0.2061); synthetic overfit persists
|
| 47 |
N2_vqce_multi_nat: TRANSFER INVERSION — best at ckpt #1 (0.1913) vs final (0.1925); synthetic overfit persists
|
| 48 |
N4_rawce_nat: TRANSFER INVERSION — best at ckpt #1 (0.2043) vs final (0.2437); synthetic overfit persists
|
|
|
|
| 28 |
## Results
|
| 29 |
```
|
| 30 |
=== transfer curve: GM-CRPS by checkpoint (dev-14, mean over seeds) ===
|
| 31 |
+
gm_avg2 gm_avg2_std gm@7500 gm@15000 gm@22500 gm@final gm_final_std long_season params_m perp probe_amp n
|
| 32 |
+
arm
|
| 33 |
+
R3_k4096_nat 0.1795 NaN 0.2066 0.1889 0.1840 0.1750 NaN 0.2512 4.2743 1597.8119 0.2750 1
|
| 34 |
+
R4_softce_nat 0.1817 0.0048 0.2018 0.1960 0.1834 0.1800 0.0099 0.2508 3.8780 558.6021 0.2380 2
|
| 35 |
+
N3_vqce_nobonus_nat 0.1835 0.0021 0.1987 0.1881 0.1883 0.1787 0.0034 0.2485 3.8780 555.6374 0.1750 3
|
| 36 |
+
R1_w01_nat 0.1836 NaN 0.1908 0.1854 0.1862 0.1809 NaN 0.2538 3.8780 570.1680 0.2180 1
|
| 37 |
+
R2_k2048_nat 0.1866 0.0012 0.2137 0.1935 0.1898 0.1834 0.0048 0.2542 4.0101 917.4004 0.2910 2
|
| 38 |
+
R5_horizonce_nat 0.1873 0.0015 0.2021 0.1873 0.1872 0.1873 0.0031 0.2519 4.0430 570.5395 0.1965 2
|
| 39 |
+
N1_vqce_nat 0.1909 0.0109 0.2121 0.2052 0.1983 0.1835 0.0049 0.2550 3.8780 620.2301 0.2280 3
|
| 40 |
+
N5_vqce_w05_nat 0.1959 NaN 0.2092 0.2005 0.1968 0.1949 NaN 0.2593 3.8780 598.1758 0.2410 1
|
| 41 |
+
N0_raw_nat 0.2037 0.0042 0.2011 0.1938 0.2014 0.2061 0.0048 0.2635 3.5802 NaN 0.2253 3
|
| 42 |
+
N2_vqce_multi_nat 0.2039 0.0189 0.1913 0.1943 0.2154 0.1925 0.0034 0.2598 4.2743 597.4680 0.2235 2
|
| 43 |
+
N4_rawce_nat 0.2404 0.0030 0.2043 0.2293 0.2371 0.2437 0.0028 0.2832 3.7452 461.1100 0.3755 2
|
| 44 |
|
| 45 |
--- tripwire checks ---
|
| 46 |
+
R3_k4096_nat vs R4_softce_nat: final gap=0.0050 -> UNRESOLVED (< 2*sigma)
|
| 47 |
+
R4_softce_nat vs N3_vqce_nobonus_nat: final gap=-0.0013 -> UNRESOLVED (< 2*sigma)
|
| 48 |
+
N3_vqce_nobonus_nat vs R1_w01_nat: final gap=0.0022 -> UNRESOLVED (< 2*sigma)
|
| 49 |
+
R1_w01_nat vs R2_k2048_nat: final gap=0.0025 -> UNRESOLVED (< 2*sigma)
|
| 50 |
+
R2_k2048_nat vs R5_horizonce_nat: final gap=0.0039 -> UNRESOLVED (< 2*sigma)
|
| 51 |
+
R5_horizonce_nat vs N1_vqce_nat: final gap=-0.0038 -> UNRESOLVED (< 2*sigma)
|
| 52 |
+
N1_vqce_nat vs N5_vqce_w05_nat: final gap=0.0114 -> UNRESOLVED (< 2*sigma)
|
| 53 |
+
N5_vqce_w05_nat vs N0_raw_nat: final gap=0.0111 -> UNRESOLVED (< 2*sigma)
|
| 54 |
N0_raw_nat vs N2_vqce_multi_nat: final gap=-0.0136 -> UNRESOLVED (< 2*sigma)
|
| 55 |
N2_vqce_multi_nat vs N4_rawce_nat: final gap=0.0513 -> RESOLVED
|
| 56 |
+
R5_horizonce_nat: TRANSFER INVERSION — best at ckpt #3 (0.1872) vs final (0.1873); synthetic overfit persists
|
| 57 |
N0_raw_nat: TRANSFER INVERSION — best at ckpt #2 (0.1938) vs final (0.2061); synthetic overfit persists
|
| 58 |
N2_vqce_multi_nat: TRANSFER INVERSION — best at ckpt #1 (0.1913) vs final (0.1925); synthetic overfit persists
|
| 59 |
N4_rawce_nat: TRANSFER INVERSION — best at ckpt #1 (0.2043) vs final (0.2437); synthetic overfit persists
|