apsua commited on
Commit
a32773b
·
verified ·
1 Parent(s): 09e2e1e

Remove unsolicited benchmark commentary

Browse files
Files changed (1) hide show
  1. README.md +0 -8
README.md CHANGED
@@ -57,8 +57,6 @@ russian = load_dataset("apsua/MIRON_Benchmark", "ru", split="train")
57
  english = load_dataset("apsua/MIRON_Benchmark", "en", split="train")
58
  ```
59
 
60
- The `train` label is the file split exposed by the dataset loader; the dataset is intended as a benchmark, so avoid training on examples used for evaluation. Select a language configuration when loading the dataset.
61
-
62
  ## Evaluation
63
 
64
  The benchmark was designed around two complementary measures:
@@ -66,12 +64,6 @@ The benchmark was designed around two complementary measures:
66
  - **Generation score:** character-level Levenshtein similarity between a generated continuation and the reference, expressed as a percentage. For two empty strings, the similarity is 100%.
67
  - **Target confidence:** geometric mean of the model probabilities assigned to the reference continuation tokens, conditioned on the prefix and preceding reference tokens, expressed as a percentage.
68
 
69
- For reproducible comparisons, report results separately by language and category, and specify the model, tokenizer, generation settings and text normalization. Token boundaries can change when `prefix` and `target` are joined; use a consistent, documented procedure for identifying target tokens. Confidence is tokenizer-dependent and should not be interpreted as a calibrated probability that an answer is correct.
70
-
71
- ## Limitations
72
-
73
- This benchmark covers a small, selected set of short continuation tasks. It does not measure general intelligence or establish the cause of an error: low scores can reflect the data, tokenizer, model or evaluation procedure. Scores are most useful when paired with task-level analysis and other evaluations.
74
-
75
  ## License
76
 
77
  MIT, as specified in the dataset metadata.
 
57
  english = load_dataset("apsua/MIRON_Benchmark", "en", split="train")
58
  ```
59
 
 
 
60
  ## Evaluation
61
 
62
  The benchmark was designed around two complementary measures:
 
64
  - **Generation score:** character-level Levenshtein similarity between a generated continuation and the reference, expressed as a percentage. For two empty strings, the similarity is 100%.
65
  - **Target confidence:** geometric mean of the model probabilities assigned to the reference continuation tokens, conditioned on the prefix and preceding reference tokens, expressed as a percentage.
66
 
 
 
 
 
 
 
67
  ## License
68
 
69
  MIT, as specified in the dataset metadata.