I’ve been a CPA for more than 40 years, and in all that time, I’ve never met an entrepreneur who likes taxes.
The BrokenMath benchmark (NeurIPS 2025 Math-AI Workshop) tested this in formal reasoning across 504 samples. Even GPT-5 produced sycophantic “proofs” of false theorems 29% of the time when the user implied the statement was true. The model generates a convincing but false proof because the user signaled that the conclusion should be positive. GPT-5 is not an early model. It’s also the least sycophantic in the BrokenMath table. The problem is structural to RLHF: preference data contains an agreement bias. Reward models learn to score agreeable outputs higher, and optimization widens the gap. Base models before RLHF were reported in one analysis to show no measurable sycophancy across tested sizes. Only after fine-tuning did sycophancy enter the chat. (literally)，详情可参考新收录的资料
，这一点在新收录的资料中也有详细论述
What actually happened was that the port to my database docker container was just out there，更多细节参见新收录的资料
was ordered and the IDPROM programming repeated. Unfortunately, this gave the same results and so more research is required, but at least the SPARCstation will keep time and it won’t be necessary to program the MAC address and hostid every time it’s powered on; not automatically booting is a minor annoyance to put up with.