samuki-hf commited on
Commit
ce134f7
·
verified ·
1 Parent(s): 5996357

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. README.md +6 -6
  2. references/data.parquet +2 -2
README.md CHANGED
@@ -17,9 +17,9 @@ tags:
17
  OCaml reference solutions for the multilingual-LiveCodeBench OCaml problem set, together
18
  with the list of problems the harness cannot score.
19
 
20
- - **415** problems in the universe
21
  - **415** are scoreable, and each has a verified reference
22
- - **0** are excluded as unscoreable (see the `errata` config)
23
 
24
 
25
  ## Configs
@@ -41,8 +41,8 @@ Every solution was compiled with `ocamlopt` and run against the real hidden test
41
  |---|---|---|
42
  | `exact-match` | passes every official test byte for byte | 410 |
43
  | `float-tolerance` | correct; differs only in float formatting | 5 |
44
- | `validated-not-exact` | proven correct; the problem accepts several answers | 0 |
45
- | `unverified` | from LiveCodeBench's errata list, not independently checked | 0 |
46
 
47
  The `float-tolerance` rows are correct solutions. Those problems accept an absolute or
48
  relative error of at most 1e-6, and no comparator in the LiveCodeBench lineage implements that
@@ -95,9 +95,9 @@ last line and a correct program fails every test.
95
  | provenance | model | solutions |
96
  |---|---|---|
97
  | model-sample | qwen3-8b-think | 239 |
98
- | agent | sonnet | 134 |
 
99
  | model-sample | qwen3-14b-think | 17 |
100
- | agent | opus | 16 |
101
  | agent | fable | 4 |
102
  | model-sample | qwen3-14b-nothink | 3 |
103
  | model-sample | llama31-8b-instruct-nothink | 1 |
 
17
  OCaml reference solutions for the multilingual-LiveCodeBench OCaml problem set, together
18
  with the list of problems the harness cannot score.
19
 
20
+ - **436** problems in the universe
21
  - **415** are scoreable, and each has a verified reference
22
+ - **21** are excluded as unscoreable (see the `errata` config)
23
 
24
 
25
  ## Configs
 
41
  |---|---|---|
42
  | `exact-match` | passes every official test byte for byte | 410 |
43
  | `float-tolerance` | correct; differs only in float formatting | 5 |
44
+ | `validated-not-exact` | proven correct; the problem accepts several answers | 12 |
45
+ | `unverified` | from LiveCodeBench's errata list, not independently checked | 8 |
46
 
47
  The `float-tolerance` rows are correct solutions. Those problems accept an absolute or
48
  relative error of at most 1e-6, and no comparator in the LiveCodeBench lineage implements that
 
95
  | provenance | model | solutions |
96
  |---|---|---|
97
  | model-sample | qwen3-8b-think | 239 |
98
+ | agent | sonnet | 153 |
99
+ | agent | opus | 17 |
100
  | model-sample | qwen3-14b-think | 17 |
 
101
  | agent | fable | 4 |
102
  | model-sample | qwen3-14b-nothink | 3 |
103
  | model-sample | llama31-8b-instruct-nothink | 1 |
references/data.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7205130235e6045f4abb5fe36a61051dbb89175b8146d63476a6f1dfc20275d8
3
- size 509822
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:895b099625255577c50a24349d4dc8f8a73105c262c414537326004c0f6c784d
3
+ size 540173