SnowFX commited on
Commit
d8cd3a3
·
verified ·
1 Parent(s): 17321fd

SimpleQA results from LocalAgentBrowser: the standard 200 (1.66, Phi-4) and six 50-question runs

Browse files
Files changed (3) hide show
  1. README.md +98 -0
  2. simpleqa_200.jsonl +0 -0
  3. simpleqa_50.jsonl +0 -0
README.md ADDED
@@ -0,0 +1,98 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - en
5
+ pretty_name: LocalAgentBrowser on SimpleQA
6
+ size_categories:
7
+ - n<1K
8
+ task_categories:
9
+ - question-answering
10
+ tags:
11
+ - simpleqa
12
+ - web-agent
13
+ - evaluation
14
+ - grounding
15
+ - citations
16
+ - small-models
17
+ configs:
18
+ - config_name: simpleqa_200
19
+ data_files: simpleqa_200.jsonl
20
+ default: true
21
+ - config_name: simpleqa_50
22
+ data_files: simpleqa_50.jsonl
23
+ ---
24
+
25
+ # LocalAgentBrowser on SimpleQA
26
+
27
+ Every answer, citation and grade from [LocalAgentBrowser](https://github.com/SnowFXXX/LocalAgentBrowser) (L.A.B.)
28
+ runs on [SimpleQA](https://openai.com/index/introducing-simpleqa), OpenAI's benchmark of short, hard factual
29
+ questions. L.A.B. is an open-source web research agent built for small models: it searches DuckDuckGo, reads pages
30
+ in a real browser, and code checks every finding against the page it came from before it counts. Its answers are
31
+ written only from those findings, each with its source.
32
+
33
+ ## Results
34
+
35
+ **`simpleqa_200`** (default): the 200 questions OpenAI's [simple-evals](https://github.com/openai/simple-evals) runs
36
+ with `--examples 200`, on L.A.B. 1.66 with Phi-4 through OpenRouter, November 16, 2026. The run took 142 minutes and
37
+ 2,762 model calls, for $0.58.
38
+
39
+ | Grader | Correct | Wrong | Not attempted | F-score |
40
+ |---|---|---|---|---|
41
+ | Claude, judgement calls reviewed by the author | **170 (85%)** | 11 | 19 | **0.892** |
42
+ | OpenAI's simple-evals grader (GPT-4.1) | 171 (85.5%) | 15 | 14 | 0.886 |
43
+
44
+ When it answered, it was right 94% of the time. The two graders give the same grade on 194 of the 200.
45
+
46
+ **`simpleqa_50`**: six runs on an earlier random 50 (seed 20261106), including the models alone with no browsing:
47
+
48
+ | Run | Correct | Wrong | Not attempted | F-score | OpenAI's grader: correct |
49
+ |---|---|---|---|---|---|
50
+ | L.A.B. 1.66, Phi-4 | 45 | 4 | 1 | 0.909 | 46 |
51
+ | L.A.B. 1.65.1, Phi-4 (Nov 15) | 39 | 5 | 6 | 0.830 | 39 |
52
+ | L.A.B. 1.65.1, Phi-4 (Nov 11) | 38 | 6 | 6 | 0.809 | 38 |
53
+ | Phi-4 alone, no browsing | 1 | 34 | 15 | 0.024 | 3 |
54
+ | L.A.B. 1.64, Qwen2.5-7B, local on a CPU (no GPU) | 35 | 6 | 9 | 0.769 | 36 |
55
+ | Qwen2.5-7B alone, no browsing | 0 | 13 | 37 | 0 | 0 |
56
+
57
+ ## Fields
58
+
59
+ | Field | What it is |
60
+ |---|---|
61
+ | `run` | Which run the row belongs to |
62
+ | `setting` | L.A.B. with a cloud or local model, or the model alone |
63
+ | `model`, `lab_version` | The model, and the L.A.B. version (none for the model alone) |
64
+ | `question_number` | The question's number in its sample (1 to 200, or 1 to 50) |
65
+ | `simpleqa_index` | The question's row in SimpleQA's `simple_qa_test_set.csv` (0-based) |
66
+ | `topic`, `question`, `reference_answer` | From SimpleQA |
67
+ | `answer` | The final answer, exactly as written, with its citation numbers |
68
+ | `cited_discoveries` | The discovery numbers the answer cites |
69
+ | `discoveries` | Every finding the agent recorded: its number, the page's URL and the text taken from the page |
70
+ | `grade` | CORRECT, INCORRECT or NOT_ATTEMPTED: graded by Claude, judgement calls reviewed by the author |
71
+ | `grade_note` | Why, for every wrong answer, judgement call and borderline grade |
72
+ | `grade_openai` | The grade from simple-evals' own grader prompt and model (GPT-4.1) |
73
+
74
+ ## How it was run and graded
75
+
76
+ - Each question was given to the agent as an ordinary objective, with 60 minutes and nothing carried over from the
77
+ question before.
78
+ - Sites that republish SimpleQA with its answers were blocked, so every answer had to be found on the open web.
79
+ - Search: DuckDuckGo's free HTML results, not a paid search API.
80
+ - Grading: SimpleQA's three grades. Claude graded each answer against the reference, one at a time, before seeing
81
+ the automatic grades. The author then reviewed every judgement call and borderline grade (48 in all; 2 changed).
82
+ OpenAI's grader is reported beside it as a second, automatic grade.
83
+ - The questions, scripts, reports and grading notes are in the repository's
84
+ [benchmarks/simpleqa](https://github.com/SnowFXXX/LocalAgentBrowser/tree/main/benchmarks/simpleqa) folder.
85
+ `run.py sample --size 200 --simple-evals` draws the same 200 from the dataset file.
86
+
87
+ ## Caveats
88
+
89
+ - Every score here is self-reported by the project that built the agent.
90
+ - 200 questions give a likely range of about 79% to 89% for 85%; 50 questions about 79% to 96% for 45 of 50.
91
+ - Phi-4 is a 14B model run in the cloud for these runs. The local run is a different model, Qwen2.5-7B, on a CPU.
92
+ - SimpleQA has been public since 2024, and some of its reference answers are known to be wrong.
93
+
94
+ ## License
95
+
96
+ MIT, like the LocalAgentBrowser repository. SimpleQA's questions and reference answers are © OpenAI under the MIT
97
+ license. The `discoveries` texts are short excerpts from the pages listed beside them, and remain under those pages'
98
+ own terms.
simpleqa_200.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
simpleqa_50.jsonl ADDED
The diff for this file is too large to render. See raw diff