vilhess commited on
Commit
682e274
·
verified ·
1 Parent(s): 7b8a0f7

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +19 -28
README.md CHANGED
@@ -23,23 +23,25 @@ Our model (with leakage) is deployed on the [TS-Arena benchmark](https://ts-aren
23
 
24
  **Our model (without leakage) is competitive with the state-of-the-art notably compared to [MOIRAI2.0](https://arxiv.org/pdf/2511.11698)**
25
 
26
- | Model Name | Win Rate | Skill Score |
27
- |----------------------|----------|-------------|
28
- | (🥇) Chronos-2 | 0.9008 | 0.4728 |
29
- | (🥈) TiRex | 0.8092 | 0.4268 |
30
- | (🥉) TimesFM-2.5 | 0.8085 | 0.4668 |
31
- | (4) Toto-1.0 | 0.7069 | 0.4110 |
32
- | (5) **PatchFM (us)** | 0.6684 | 0.3984 |
33
- | (6) Moirai-2.0 | 0.6546 | 0.4026 |
34
- | (7) Chronos-Bolt | 0.6277 | 0.3889 |
35
- | (8) Sundial-Base | 0.4446 | 0.3387 |
36
- | (9) CatBoost (Recursive)| 0.3362 | 0.2301 |
37
- | (10) LightGBM (Recursive)| 0.3123 | 0.2168 |
38
- | (11) AutoTheta | 0.2938 | 0.0546 |
39
- | (12) Seasonal Naive | 0.2058 | 0.0000 |
40
- | (13) Naive | 0.1404 | -0.4540 |
41
- | (14) Drift | 0.0915 | -0.4578 |
42
-
 
 
43
 
44
  ## Highlights
45
  - Next-patch prediction objective (autoregressive, causal)
@@ -161,17 +163,6 @@ A common workaround is to feed only the median prediction (the 0.5 quantile) bac
161
 
162
  An alternative approach is **autoregressive multi-quantile decoding**, as proposed in [Moirai 2.0](https://arxiv.org/pdf/2511.11698v1). This method enables consistent autoregressive generation while preserving the full predictive distribution across quantiles. However, it is computationally more expensive than the median-only approach as it requires duplicating the context for each quantile.
163
 
164
- <div style="display: flex; gap: 10px; align-items: flex-start;">
165
- <div>
166
- <img src="assets/classic_forecast_animation.gif" alt="Classic Autoregressive Inference" width="450">
167
- <p style="text-align:center;">Classic Autoregressive Inference</p>
168
- </div>
169
- <div>
170
- <img src="assets/quantile_forecast_animation.gif" alt="Autoregressive Multi-Quantile Decoding" width="450">
171
- <p style="text-align:center;">Autoregressive Multi-Quantile Decoding</p>
172
- </div>
173
- </div>
174
-
175
  The algorithm proceeds as follows:
176
 
177
  1. **Initialization**
 
23
 
24
  **Our model (without leakage) is competitive with the state-of-the-art notably compared to [MOIRAI2.0](https://arxiv.org/pdf/2511.11698)**
25
 
26
+ | Model | Win Rate | Skill Score |
27
+ |---------|----------:|------------:|
28
+ | Chronos-2 | 0.89 | 0.47 |
29
+ | TimesFM-2.5 | 0.81 | 0.47 |
30
+ | TiRex | 0.81 | 0.43 |
31
+ | FlowState | 0.71 | 0.42 |
32
+ | Toto-1.0 | 0.70 | 0.41 |
33
+ | **PatchFM** | 0.69 | 0.40 |
34
+ | Moirai-2.0 | 0.65 | 0.40 |
35
+ | Chronos-Bolt | 0.62 | 0.39 |
36
+ | TFT | 0.47 | 0.32 |
37
+ | Sundial-Base | 0.43 | 0.34 |
38
+ | PatchTST | 0.42 | 0.30 |
39
+ | CatBoost | 0.32 | 0.23 |
40
+ | LightGBM | 0.29 | 0.21 |
41
+ | AutoTheta | 0.28 | 0.05 |
42
+ | Seasonal Naive | 0.19 | 0.00 |
43
+ | Naive | 0.13 | -0.45 |
44
+ | Drift | 0.09 | -0.46 |
45
 
46
  ## Highlights
47
  - Next-patch prediction objective (autoregressive, causal)
 
163
 
164
  An alternative approach is **autoregressive multi-quantile decoding**, as proposed in [Moirai 2.0](https://arxiv.org/pdf/2511.11698v1). This method enables consistent autoregressive generation while preserving the full predictive distribution across quantiles. However, it is computationally more expensive than the median-only approach as it requires duplicating the context for each quantile.
165
 
 
 
 
 
 
 
 
 
 
 
 
166
  The algorithm proceeds as follows:
167
 
168
  1. **Initialization**