After two years of running RAG in production, here are the chunking strategies I actually use, the ones I dropped, and why chunk size is…
A new attack exploits the fact that most LLM safety layers evaluate turns in isolation. Here's what the Transient Turn Injection result means for anyone…
Why LLMs fail at spatial tasks isn't always about model size. The input format matters more than you'd expect, and here's what to do about…
New numbers on iterative LLM self-repair: where extra tries earn their keep, where they quietly waste tokens, and the retry loop I'll actually defend.
A new paper tested 8 reasoning models with 14 formatting perturbations. Open-weight models lost up to 55% accuracy. Here's what that means for production use.
A practical look at which new open-weight LLMs you can actually run on a single workstation with llama.cpp in 2026, and which ones are just…
New research scores 263 tasks across four LLMs. Math and coding top the automation list, but the jobs most exposed to AI demand the opposite…
New research shows LLM API testing misses how chatbots actually behave. Here's what the data says about the gap and how to fix your eval…
LLMs claim massive context windows but choke on long number sequences. A training-free trick called SepSeq fixes attention dispersion and boosts accuracy 35%.
A new benchmark shows LLMs drop 0.3 to 5.9% accuracy on grade-school math when names and places go non-Western. Same arithmetic, different answers.