v0.63.0
Questa è la nota di release così com'è stata pubblicata, non una sua riscrittura. Le note di release sono pubblicate nella lingua in cui sono state scritte.
Added
-
Answers grounded in your documents are checked against them, on by default. When a message asks a question about attached documents, and the answer is written from them with no tool call in the turn, a System One verifier reads question, sources and answer.
- What happens next:
- If the answer is supported with confidence ≥ 0.8, it ships.
- Otherwise it escalates to a stronger model (
gpt-6-sol), which is verified again. - If that also fails, you get "the sources provided don't cover this" in your language, instead of an unsupported answer.
- Measured in
bench/verified_cascadeon 400 grounded questions, withgpt-6-lunadrafting: wrong answers shipped went from 33 to 21 (Holm p = 0.002, 11 fixed and 0 broken), at 1.87× the cost with the local verifier, and with no answerable question handed off. - Not applied to tasks on documents (summarize, critique, translate, rewrite), to tool-using turns or to
solve. A deterministic classifier routes tasks past the check. - Where to see it: a badge on the answer (verified, sources don't cover it, lexical only, verifier unavailable, not checked), with the withheld text one click away, and the receipt's
groundedblock records every step. - Controls:
CHIMERA_VERIFIED_ANSWERS,CHIMERA_VERIFIED_ANSWERS_THRESHOLDandCHIMERA_VERIFIED_ANSWERS_ESCALATE_MODEL. - In the terminal,
chatandassistgain/attach <file>. (#656)
- What happens next:
-
Choose the System One decision model in Settings and in the terminal. A new "System One" card offers three backends: Local (qwen3:4b, free), Hosted judge, or OpenRouter System One.
- Model list: it comes live from OpenRouter's
output_modalities=decisionsindex, with price, context and a calibrated or uncalibrated badge. - Selectable today:
typesafe/jev-1.13andjaredpalmer/kev-4b. - Refused, with the reason: the moving alias
~typesafe/jev-latest, andrespan/span-01*, whose behavior-scoring contract has not been measured. - Terminal:
chimera decisions modelsandchimera decisions use <backend> [model]. - Fallback: the verified-answers check uses this choice. The local verifier falls back to Jev when an OpenRouter key exists, and to the lexical check otherwise. (#653)
- Model list: it comes live from OpenRouter's
Benchmarks
-
bench/verified_cascade— pre-registered, US$ 4.89 spent.arm wrong answers shipped (of 400) result cost Decision-gated cascade, Jev verifier 21 significant against luna alone 3.74× Decision-gated cascade, local verifier 21 significant; tied with Jev, so the local one is preferred 1.87× Today's lexical gate 22 null 10.8× gpt-6-solalone20 null 19.4× The run needed three amendments: a grader swapped after failing its gate, 42 disagreements adjudicated, and a report bug that read zero hand-offs as missing. (#654, #655)