Training a 'is this joke funny' model is hard; targeted critique prompts work better
ivan_bezdomny · x · 2026-11-11
The author found that training a decision model for "does this joke work" is challenging: asking Opus "is it funny?" doesn't help, but asking it to flag stale/cliché jokes or evaluate originality and structure does work. They also note GPT censors many jokes while Claude will readily write mean, dark ones—just not as aggressively as Grok.
Related event: Tests show Claude tells edgy jokes freely while GPT censors more(2 posts)→
More from Models
- Apple reportedly building Harbor app to turn Macs into private AI inference servers with Thunderbolt 5 clustering — BenBajarin · 2026-11-11
- Agent Arena: Gemini 4 Carbon Gets More Cost-Competitive as Tasks Get Harder — rseroter · 2026-11-11
- LiquidAI's 1.7B LFM3 hits SOTA on Multi-IF, beating 100x-larger rivals including ERNIE 5.0 — NielsRogge · 2026-11-11
- Fable 5.2 tops GDPx chip-design benchmark, but task authors see zero gains — ycombinator · 2026-11-11
- 1.3B open-source MiniCPM-OCR 1.0 tops Real5-OmniDocBench, beating GPT-5.2 — NielsRogge · 2026-11-11
- Claude writes dark, offensive jokes freely while GPT censors them, per hands-on test — ivan_bezdomny · 2026-11-11