Update README.md
Browse files
README.md
CHANGED
|
@@ -31,6 +31,34 @@ Additionally, attempted to implement MARA from https://im-ant.github.io/mara/ on
|
|
| 31 |
| piqa | acc_norm | 0.7807 | 0.7807 | ±0.00% | 0.7813 | +0.08% | +0.08% | 0.7807 | ±0.00% | -0.08% |
|
| 32 |
|
| 33 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
|
| 35 |
|
| 36 |
## Merge Details
|
|
|
|
| 31 |
| piqa | acc_norm | 0.7807 | 0.7807 | ±0.00% | 0.7813 | +0.08% | +0.08% | 0.7807 | ±0.00% | -0.08% |
|
| 32 |
|
| 33 |
|
| 34 |
+
Some very interesting results on diversity also:
|
| 35 |
+
|
| 36 |
+
**Diversity Metrics (Qwen3-4B-Base vs Style-Karcher, temperature=1.0, 8 completions per prompt)**
|
| 37 |
+
|
| 38 |
+
| Domain | Metric | Base | Karcher | Δ |
|
| 39 |
+
|--------|--------|:----:|:-------:|:-:|
|
| 40 |
+
| ao3_english | Prefix entropy | 3.309 | 3.238 | -2.1% |
|
| 41 |
+
| ao3_english | Distinct-1 | 0.618 | **0.683** | **+10.5%** |
|
| 42 |
+
| ao3_english | Distinct-2 | 0.962 | **0.984** | +2.3% |
|
| 43 |
+
| ao3_english | Pairwise diversity | 0.919 | **0.932** | +1.4% |
|
| 44 |
+
| github_python | Prefix entropy | 1.514 | 1.456 | -3.8% |
|
| 45 |
+
| github_python | Distinct-1 | 0.610 | **0.624** | +2.3% |
|
| 46 |
+
| github_python | Distinct-2 | 0.890 | 0.876 | -1.6% |
|
| 47 |
+
| github_python | Pairwise diversity | 0.933 | 0.933 | ±0.0% |
|
| 48 |
+
| wikipedia_english | Prefix entropy | 1.974 | 1.892 | -4.2% |
|
| 49 |
+
| wikipedia_english | Distinct-1 | 0.599 | 0.559 | -6.7% |
|
| 50 |
+
| wikipedia_english | Distinct-2 | 0.932 | 0.898 | -3.6% |
|
| 51 |
+
| wikipedia_english | Pairwise diversity | 0.907 | 0.900 | -0.8% |
|
| 52 |
+
| bbc_news | Prefix entropy | 2.252 | 2.186 | -2.9% |
|
| 53 |
+
| bbc_news | Distinct-1 | 0.557 | **0.577** | +3.6% |
|
| 54 |
+
| bbc_news | Distinct-2 | 0.949 | **0.951** | +0.3% |
|
| 55 |
+
| bbc_news | Pairwise diversity | 0.901 | **0.908** | +0.8% |
|
| 56 |
+
| arxiv_cs | Prefix entropy | 2.455 | 2.346 | -4.4% |
|
| 57 |
+
| arxiv_cs | Distinct-1 | 0.555 | **0.567** | +2.3% |
|
| 58 |
+
| arxiv_cs | Distinct-2 | 0.905 | **0.906** | +0.2% |
|
| 59 |
+
| arxiv_cs | Pairwise diversity | 0.895 | **0.901** | +0.7% |
|
| 60 |
+
|
| 61 |
+
|
| 62 |
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
|
| 63 |
|
| 64 |
## Merge Details
|