memesearch
Offensive Cyber Evaluations Chart

Offensive Cyber Evaluations Chart

Download Find similar pictures
Text on the meme
Offensive cyber evaluations Claude Opus 4.8 (no safeguards) Claude Opus 4.8 (default safeguards) Claude Mythos Preview Claude Mythos 5 Claude Fable Success rate (%) 100 80 60 40 20 0 88.4 70.8 8.8 0.0 15.9 22.8 24.0 3.8 0.0 78.1 83.1 83.8 0.8 0.0 n/d 29.2 38.7 0.0 Firefox OSS-Fuzz CyberGym CyScenarioBench
What it means

A bar chart compares Claude model variants on offensive cybersecurity evaluations across Firefox, OSS-Fuzz, CyberGym, and CyScenarioBench. The joke is that the supposedly lobotomized or safeguarded models score much higher than the safeguarded baseline, while Claude Fable scores zero or is unavailable, exaggerating the idea that safety restrictions cripple performance.

Objects
bar chart / столбчатая диаграммаdata bars / столбцы данныхFirefox label / подпись FirefoxOSS-Fuzz label / подпись OSS-FuzzCyberGym label / подпись CyberGymCyScenarioBench label / подпись CyScenarioBenchlegend / легендаgreen bars / зелёные столбцыblue bars / синие столбцыpink bars / розовые столбцыorange bars / оранжевые столбцыpercentage axis / процентная осьgrid lines / линии сеткиwhite background / белый фон
Tags
#cybersecurity#кибербезопасность#ai#ии#claude#anthropic#mythos#fable#safety#безопасность#benchmark#бенчмарк#bar chart#диаграмма#reaction#реакция#absurd#абсурд#tech meme#техномем#safeguards#защитные ограничения
Source
Type
image
Size
800×450 · 36.1 KiB
Mood
sarcastic, absurd
Language
en
Added

Similar memes

o3 Crushes Coding Benchmarks o3 Crushes Coding Benchmarks AI Felony Bench Scorecard AI Felony Bench Scorecard Opus 5 Benchmark Sweep Opus 5 Benchmark Sweep AI Benchmark Comparison Table AI Benchmark Comparison Table Opus Dominates Long-Context Retrieval Opus Dominates Long-Context Retrieval Opus 4.6 Benchmark Sweep Opus 4.6 Benchmark Sweep Misleading AI Accuracy Graph Misleading AI Accuracy Graph Opus 4.1 SWE-bench Accuracy Opus 4.1 SWE-bench Accuracy More Compute, Mixed Results More Compute, Mixed Results o3 Beats Humans at Math and Science o3 Beats Humans at Math and Science GPT-5 Numbers Go Sky-High GPT-5 Numbers Go Sky-High Gemini 2.5 Pro Benchmark Table Gemini 2.5 Pro Benchmark Table Trust in Copy and Paste Trust in Copy and Paste AI Model Name Escalation Chart AI Model Name Escalation Chart Humanity's Last Exam Results Humanity's Last Exam Results GPT-5.3-Codex Benchmark Flex GPT-5.3-Codex Benchmark Flex Humanity’s Last Exam Results Humanity’s Last Exam Results BullshitBench Model Improvement Chart BullshitBench Model Improvement Chart GPT Tool Use Benchmark Chart GPT Tool Use Benchmark Chart o3 Research Math Accuracy Chart o3 Research Math Accuracy Chart TOON Benchmark Accuracy Meme TOON Benchmark Accuracy Meme Leading Causes of Death Chart Leading Causes of Death Chart Me Versus Your Exes IQ Chart Me Versus Your Exes IQ Chart GitHub Average Uptime Collapse GitHub Average Uptime Collapse Cloudflare vs Vercel CPU Benchmarks Cloudflare vs Vercel CPU Benchmarks Preferred Code Editor Poll Preferred Code Editor Poll Распределение IQ и тесты кода Распределение IQ и тесты кода SWE-Bench Pro Accuracy Chart SWE-Bench Pro Accuracy Chart Yandex Contributes 44 Gb Yandex Contributes 44 Gb How Men and Women Rate Attractiveness How Men and Women Rate Attractiveness Дефицит медицинских масок Дефицит медицинских масок Kimi-K3 Tops the Coding Arena Kimi-K3 Tops the Coding Arena 2024 Internet Traffic Breakdown 2024 Internet Traffic Breakdown Programmers Identifying as LGBTQ by Language Programmers Identifying as LGBTQ by Language Claude Models Playing Pokémon Claude Models Playing Pokémon Firewall Backdoor Protection Chain Firewall Backdoor Protection Chain