Memcached - phpCacheAdmin

PHPMem v2.0.1

Version

1.6.45

Uptime

7 days 14 hours 45 minutes 46 seconds

Memory

Total

512MB

Used

13,23MB (2.58%)

Free

498,77MB

Keys

Current

10 092

Total (since start)

11 096

Evictions

Reclaimed

190

Expired Unfetched

Evicted Unfetched

Connections

Current

4 / 1 024 max

Total

69 801

Rejected

Key

Expire (in seconds) 0 removes expiration (default).

Encode function

Value

{"action": "final", "content": "## Most Common Values Across Main Categorical Columns\n\nBased on the analysis of the benchmarks dataset, here are the most frequent values for each main categorical dimension:\n\n### Display Name (Model Identifiers)\nThe dataset contains **309 blank/missing display names (23.57%)**, which is the most common value. Among named models:\n- **Gemini 3.5 Flash**: 20 occurrences (1.53%)\n- **Gemini 3.1 Pro Preview**: 18 occurrences (1.37%)\n- **Claude Opus 4.8**: 17 occurrences (1.30%)\n- **Gemini 3 Flash Preview**: 17 occurrences (1.30%)\n\nThe distribution is highly fragmented across ~177 distinct display names, with no single model dominating beyond the missing values.\n\n### Domain (Model Capabilities)\n- **Language**: 449 occurrences (34.25%) — pure language models are the most common\n- **Multimodal,Language,Vision**: 304 occurrences (23.19%)\n- **Language,Multimodal,Vision**: 180 occurrences (13.73%)\n- **Language,Vision,Multimodal**: 106 occurrences (8.09%)\n- **Blank/missing**: 86 occurrences (6.56%)\n\nLanguage-focused models dominate, with various multimodal combinations making up most of the remainder.\n\n### Organization (Model Developers)\n- **OpenAI**: 428 occurrences (32.65%) — the clear leader\n- **Anthropic**: 296 occurrences (22.58%)\n- **Google DeepMind**: 132 occurrences (10.07%)\n- **Alibaba**: 131 occurrences (9.99%)\n- **DeepSeek**: 52 occurrences (3.97%)\n\nOpenAI and Anthropic together account for over half (55%) of all benchmark runs.\n\n### Model (Technical Identifiers)\nThe top models by frequency:\n- **gemini-3.5-flash_high**: 13 occurrences (0.99%)\n- **gpt-5-2025-08-07_high**: 13 occurrences (0.99%)\n- **claude-opus-4-6_max**: 12 occurrences (0.92%)\n- **kimi-k2.6**: 12 occurrences (0.92%)\n\nWith ~325 distinct model identifiers, the distribution is extremely fragmented — no single model exceeds 1% of the dataset.\n\n### Task (Benchmark Tests)\n- **GPQA diamond**: 255 occurrences (19.45%) — the most frequently run benchmark\n- **OTIS Mock AIME 2024-2025**: 230 occurrences (17.54%)\n- **Chess Puzzles**: 153 occurrences (11.67%)\n- **MATH level 5**: 108 occurrences (8.24%)\n- **FrontierMath-2025-02-28-Private**: 101 occurrences (7.70%)\n\nThe top 5 tasks account for about 65% of all benchmark runs, with GPQA and OTIS AIME being the most popular evaluation targets.\n\n### Key Observations\n- **High fragmentation**: Model identifiers are highly diverse (325 distinct models, 177 display names), with no single model dominating\n- **Organizational concentration**: OpenAI and Anthropic together represent 55% of runs\n- **Task focus**: A small set of challenging benchmarks (GPQA, AIME, Chess, MATH) drive most evaluations\n- **Missing data**: 23.6% of display names are blank, suggesting incomplete metadata for nearly a quarter of runs", "charts": [{"kind": "hbar", "x": "value", "series": ["count"], "title": "Top 10 Display Names by Frequency"}, {"kind": "bar", "x": "value", "series": ["count"], "title": "Domain Distribution"}, {"kind": "bar", "x": "value", "series": ["count"], "title": "Organization Distribution"}, {"kind": "bar", "x": "value", "series": ["count"], "title": "Task Distribution"}]}