← eval history

Qwen/Qwen3-Embedding-8B + deepseek-chat · k=20

run #32 · 2026-07-21 01:04 · git be52b65

Retrieval hit@k
98%
Answer accuracy
96%
Cases
45 (40 grounded)

Δ vs baseline

Compared to Qwen/Qwen3-Embedding-8B + deepseek-chat · k=20 (2026-07-20 11:20). A retrieval delta with the same generator isolates the embedder/retrieval change.

Failures in this run (3)

CaseQuestionRetrievalAnswerSymptom
q014By how much did Berkshire Hathaway's insurance float grow from 2022 to 2023?PASSFAILretrieved but generated wrong (generation/arithmetic)
q024How much larger were Berkshire Hathaway's 2023 net earnings than its 2023 operating earnings?PASSFAILretrieved but generated wrong (generation/arithmetic)
q043Which had the highest net income in 2023: JPMorgan Chase, Microsoft, or Costco?FAILPASSanswered without the gold chunk (parametric leak?)