The only reasoning failures here are in the minds of humans gulled into expecting chatbot reasoning ability.

Papers like these are much needed bucket of ice water. We antropomorphize these systems too much.

Skimming through conclusions and results, the authors conclude that LLMs exhibit failures across many axes we'd find to be demonstrative of AGI. Moral reasoning, simple things like counting that a toddler can do, etc. They're just not human and you can reasonably hypothesize most of these failures stem from their nature as next-token predictors that happen to usually do what you want.

So. If you've got OpenClaw running and thinking you've got Jarvis from Iron Man, this is probably a good read to ground yourself.

Note there's a GitHub repo compiling these failures from the authors: https://github.com/Peiyang-Song/Awesome-LLM-Reasoning-Failur...