Add OB-News-Websearch eval results (private 300)
Browse files
.eval_results/ob-news-websearch.yaml
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
- dataset:
|
| 2 |
+
id: openbenchmarks/OB-News-Websearch
|
| 3 |
+
task_id: accuracy
|
| 4 |
+
value: 93.3
|
| 5 |
+
date: '2026-09-01'
|
| 6 |
+
source:
|
| 7 |
+
url: https://openbenchmarks.com/company-news
|
| 8 |
+
name: OpenBenchmarks company-news
|
| 9 |
+
org: openbenchmarks
|
| 10 |
+
notes: "POST /v1/search · mode=basic; $ / 1k correct=$5.36; private held-out 300; extract gpt-5.6-terra medium; judge claude-opus-5 Bedrock"
|
| 11 |
+
- dataset:
|
| 12 |
+
id: openbenchmarks/OB-News-Websearch
|
| 13 |
+
task_id: web_search
|
| 14 |
+
value: 93.3
|
| 15 |
+
date: '2026-09-01'
|
| 16 |
+
source:
|
| 17 |
+
url: https://openbenchmarks.com/company-news
|
| 18 |
+
name: OpenBenchmarks company-news
|
| 19 |
+
org: openbenchmarks
|
| 20 |
+
notes: "POST /v1/search · mode=basic; $ / 1k correct=$5.36; private held-out 300; extract gpt-5.6-terra medium; judge claude-opus-5 Bedrock"
|
| 21 |
+
- dataset:
|
| 22 |
+
id: openbenchmarks/OB-News-Websearch
|
| 23 |
+
task_id: ar_at_1
|
| 24 |
+
value: 55.0
|
| 25 |
+
date: '2026-09-01'
|
| 26 |
+
source:
|
| 27 |
+
url: https://openbenchmarks.com/company-news
|
| 28 |
+
name: OpenBenchmarks company-news
|
| 29 |
+
org: openbenchmarks
|
| 30 |
+
notes: "POST /v1/search · mode=basic; $ / 1k correct=$5.36; private held-out 300; extract gpt-5.6-terra medium; judge claude-opus-5 Bedrock"
|
| 31 |
+
- dataset:
|
| 32 |
+
id: openbenchmarks/OB-News-Websearch
|
| 33 |
+
task_id: ar_at_5
|
| 34 |
+
value: 92.0
|
| 35 |
+
date: '2026-09-01'
|
| 36 |
+
source:
|
| 37 |
+
url: https://openbenchmarks.com/company-news
|
| 38 |
+
name: OpenBenchmarks company-news
|
| 39 |
+
org: openbenchmarks
|
| 40 |
+
notes: "POST /v1/search · mode=basic; $ / 1k correct=$5.36; private held-out 300; extract gpt-5.6-terra medium; judge claude-opus-5 Bedrock"
|