Surprisingly hard to finetune models to avoid unnecessary hyphens
ivan_bezdomny · x · 2026-08-22
Attempts to stop Claude from using unnecessary hyphens face difficulties. While finetuning can remove all dashes, it leads to incorrect formatting in headlines like "GPT 5.4" or "voted 7 2", highlighting the complexity of correcting model behavior.
More from Models
- Grok 4.6 Launches on Google Vertex AI with 500k Context — SpaceXAI · 2026-08-22
- OpenAI Cuts GPT-5.6 Sol Pricing by 20% for 3 Months — Angaisb_ · 2026-08-22
- Testing H3 T2V: struggles with eye color fidelity at medium distance — SIR_NVAX_A_LOT · 2026-08-22
- Ox Alpha Speculated to be Microsoft's Unreleased MAI-2 Frontier Model — inductionheads · 2026-08-22
- User review: Anthropic's Mythos model disappoints on hype — haider1 · 2026-08-22
- Qwen3-8B Case Study: Picking First Option Then Fabricating Reasoning — a_karvonen · 2026-08-22