You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit ccc03c9
Browse filesBrowse the repository at this point in the historyBrowse files
fix(answer-commit): keep the rounded scale of recognized float noise
- A recognized float artifact now contributes the scale of the shortest value it is an artifact of, so a column mixing it with coarser cells keeps the intended digits.
- A column containing only noise cells keeps its values as given.
- Add a regression test.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Copy file name to clipboardExpand all lines: packages/opencode/src/altimate/tools/answer-commit.ts
+22-17Lines changed: 22 additions & 17 deletions
Original file line number
Diff line number
Diff line change
@@ -217,29 +217,34 @@ export function renderAnswer(payload: unknown): string {
217
217
// Float noise: more than 6 decimals, yet either a double-precision artifact of, or the exact float32 widening of, a value with at most 6 decimals (0.30000000000000004, 18.440000534057617). Genuine long decimals (1234567.8901234) are not.
218
218
// Two units in the last place of a double: arithmetic artifacts (0.1 + 0.2) sit within it, genuine 15+ digit decimals do not.
0 commit comments