4 posts tagged failure-modes.
Driving Chrome in an agentic loop, Qwen 3.8 Flash-Next navigates to a signed Alibaba Cloud object-storage URL it invented: 235 times in 17,679 navigations. Under identical workloads and prompts, Claude, GPT-5.6 Sol, GLM-5.3 and DeepSeek do not. It also reports sending such URLs when it did not.
vLLM's Triton top-k/top-p kernel drops the weakest surviving candidate when min_p, top_k and top_p are used together. Nine days of runs decoded near-greedily, every correctness instrument passed, and the investigation started on the weights.
The same model scored 11 points lower on vLLM's Anthropic bridge than on llama.cpp. The bridge hands the model its own prior reasoning back, it imitates itself, and within a few hundred steps 40% of its thinking is one 62-character sentence.
Transcript autopsy of every run that stopped without finishing: a model that mimicked its own memory system's voice, a promised file that never appeared, and the harness semantics that let both count as 'done.'