Remove checkpoint availability section from model card
Browse files
README.md
CHANGED
|
@@ -88,8 +88,6 @@ The evaluation records contain 2,417 MuSiQue examples, 132 GSM-Hard examples, 50
|
|
| 88 |
|
| 89 |
CL is best or joint-best on **6 of 10 metrics**, improves over Qwen3-14B on all ten, and gains an average of **2.38 percentage points** over that single-model reference. On the MATH evaluation, CL scores **392/500 correct**, or **78.40% accuracy and 85.25% F1**, under the evaluation scorer.
|
| 90 |
|
| 91 |
-
The exact MultiPL-E 8B checkpoint used for the reported result is unavailable; see [checkpoint availability](#checkpoint-availability) below.
|
| 92 |
-
|
| 93 |
### 📉 Efficiency and task performance
|
| 94 |
|
| 95 |
Figure 4 compares task performance with parameter-weighted output cost and average model invocations across the five benchmarks.
|
|
@@ -144,10 +142,6 @@ The release contains **233 surviving checkpoints across 29 model-stage groups**,
|
|
| 144 |
|
| 145 |
Weights are byte-for-byte unchanged. Adapter configurations use the corresponding `Qwen/Qwen3-*` base-model IDs, and JSON metadata has local server paths normalized and credential fields removed. Baseline, ablation, and homogeneous-model weights are outside this release, as are optimizer/RNG states, pickled training arguments, and training data.
|
| 146 |
|
| 147 |
-
### Checkpoint availability
|
| 148 |
-
|
| 149 |
-
The source server's 2026-11-05 cleanup deleted 18 checkpoints from the selected runs. In particular, **MultiPL-E 8B `checkpoint-25`, used in the paper's final evaluation, could not be recovered from either source server**. The same run's `checkpoint-100` and `final` are included as surviving exports; they are not substitutes for checkpoint 25 and are not assigned its reported scores. The manifest lists the known deletions. Earlier retention limits may also have removed checkpoints.
|
| 150 |
-
|
| 151 |
## 🚀 Quickstart
|
| 152 |
|
| 153 |
### 1️⃣ Install the loading dependencies
|
|
|
|
| 88 |
|
| 89 |
CL is best or joint-best on **6 of 10 metrics**, improves over Qwen3-14B on all ten, and gains an average of **2.38 percentage points** over that single-model reference. On the MATH evaluation, CL scores **392/500 correct**, or **78.40% accuracy and 85.25% F1**, under the evaluation scorer.
|
| 90 |
|
|
|
|
|
|
|
| 91 |
### 📉 Efficiency and task performance
|
| 92 |
|
| 93 |
Figure 4 compares task performance with parameter-weighted output cost and average model invocations across the five benchmarks.
|
|
|
|
| 142 |
|
| 143 |
Weights are byte-for-byte unchanged. Adapter configurations use the corresponding `Qwen/Qwen3-*` base-model IDs, and JSON metadata has local server paths normalized and credential fields removed. Baseline, ablation, and homogeneous-model weights are outside this release, as are optimizer/RNG states, pickled training arguments, and training data.
|
| 144 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
## 🚀 Quickstart
|
| 146 |
|
| 147 |
### 1️⃣ Install the loading dependencies
|