#llm

Articles tagged with #llm

LLM-as-a-Judge Bias: Why Your Eval Score Climbs While Your Answers Get Worse
Position, length, and self-preference bias in LLM judges, with the swap test, the pad test, and the calibration code to catch them
Oct 19, 202611 min read
Your LLM's Explanation Isn't Its Reason: How to Test What Actually Drove the Output
Self-explanations are generated after the fact, from the transcript. Here's the research, where it breaks real systems, and a small harness to find the real cause.
Oct 18, 202611 min read
How to Upgrade Your LLM Without Breaking Your Prompts
Old prompt patches over-correct on newer models. A spec/patch ledger, two-sided tests, and a five-step upgrade drill.
Oct 15, 20269 min read
Prompt Caching for AI Agents: Stop Breaking Your Own Cache
LLM APIs re-read the whole conversation on every call. How the prefix cache works, the bugs that silently switch it off, and the append-only rule that fixes them.
Oct 11, 202610 min read1
Structured Outputs Still Hallucinate: Designing LLM Schemas That Can Say "I Don't Know"
Constrained decoding guarantees valid JSON. It says nothing about whether the values are true.
Sep 29, 20267 min read
Temperature 0 Won't Fix Your AI. It Just Hides the Doubt.
Why LLM nondeterminism is a signal, not a bug, and how to use it
Sep 25, 20267 min read