Kimi K2 Thinking
PulseAugur coverage of Kimi K2 Thinking — every cluster mentioning Kimi K2 Thinking across labs, papers, and developer communities, ranked by signal.
-
AI safety experiment tests LLM memory instruction adherence with larger user profiles
An AI safety experiment explored how the size of a user profile affects an LLM's adherence to instructions regarding memory usage. The study partially reproduced findings from the PersistBench paper, which investigates …
-
Claude Code mixes Kimi K2 Thinking output into responses
A user reported an unusual incident where Claude Code, specifically the Sonnet 5 model, unexpectedly included output from Kimi K2 Thinking within its response. This occurred during a session where the user was asking se…
-
LLMs struggle to delete code, hindering maintainability, new research finds
A new research paper identifies "deletion avoidance" as a key issue in large language models' code editing capabilities, where models tend to retain code that should be removed. This behavior, observed across leading mo…
-
AI Safety Research Pushes for Model Forensics to Uncover Intent
Researchers are advocating for increased focus on "model forensics," a field dedicated to investigating the root causes of concerning AI behavior. The core idea is that simply observing a negative action from a model is…
-
Nvidia Rubin GPU promises 10x cheaper tokens but with significant caveats
Nvidia has announced its Vera Rubin NVL72 GPU, promising up to a 10x reduction in cost per token compared to its Blackwell architecture. However, this significant cost saving is contingent on several factors, including …
-
Chinese AI Labs Release Frontier Models Qwen 3.5, GLM 5, and MiniMax 2.5
Several Chinese AI labs have released new flagship open-weight models, including Qwen 3.5, GLM 5, and MiniMax 2.5. These releases represent a significant push in the frontier of AI development from these organizations. …