GPT-5.6 Pro still gets stuck on the Jacobian Conjecture
tokenbender · x · 2026-07-20
The post says this is already the fourth example of someone trying many manual approaches with GPT-5.6 Pro and still failing, suggesting the model’s competence is jagged rather than uniform.
In the quoted thread, the author says they ran six separate GPT-5.6 Pro instances, all without search, and none could disprove the Jacobian Conjecture. Even with hints and reminders to widen the search space, the model kept looking in an extremely narrow region of polynomial families and got stuck.
The takeaway is that for hard math problems, the model can easily miss the right search strategy, and the poster recommends combining multiple models instead of relying on one system alone.
More from Models
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- OpenAI’s Codex + GPT-5.6 Sol hits 99% recall in Project APE verification tests — soumitrashukla9 · 2026-07-22
- OpenAI rolls out voice in GPT-Live, but the UI obscures search and reasoning — Graham_dePenros · 2026-07-22
- Moonshot’s Kimi K3 sets a new open-weights ECI record at 156 — scaling01 · 2026-07-22
- Nanbeige4.2-3B launches as a 3B Looped Transformer model that beats larger baselines — Wooden-Deer-1276 · 2026-07-22
- A post says six companies now beat Google’s best LLM, including two open-source models — soham_btw · 2026-07-22