[qwist1233-cpu]

Description

scoreVerdict() in ObjectiveHeader.tsx compared score_value against the literal "true" without looking at score_type, so every nonboolean score with a value became Failure in the Attack Result Details popover: Score(score_type="unknown", score_value="unknown") read Failure, and so did a float_scale score of 0.9. The message score details already show the raw value for the same score, so the two views disagreed.

Only true_false carries a verdict (Score.get_value() returns a bool for it and a float for float_scale, and _check_score_value() restricts true_false values to true/false). The display helper now follows that: a true_false score still reads Success/Failure, every other type shows the value the scorer recorded, a score without a value stays Undetermined, and no score at all stays Not set. No numeric value is interpreted, so there is no threshold in the presentation layer. The human-score radio group reads the same helper, so a nonboolean automated score no longer preselects Failure for the reviewer either.

Closes #3916

Tests and Documentation

Added a parameterized case to ObjectiveHeader.test.tsx covering float_scale 0 and 0.9, unknown "unknown", and unknown "true" — the literal that used to produce a verdict from the wrong type. Each case asserts the raw value is shown, that the automated-score row contains no Success/Failure/Undetermined text, that the button's accessible name matches, and that the human verdict is not preselected to Failure. The existing distinguishes automated score cases keep covering true_false True/False, the valueless Undetermined score and the missing Not set score.

Verified with npx jest src/components/Chat/ObjectiveHeader.test.tsx in frontend/. The four new cases fail on main (they read Failure) and pass with this change; the file's other tests pass unchanged. No documentation change: this is a display fix with no API surface.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

No active review issues are reported for the change to display raw nonboolean scores in Attack Result Details.

No open issues 🎉

What changed in this PR

Fixes #3916 by showing raw nonboolean scores in Attack Result Details instead of incorrectly labeling them as Success or Failure, aligning this view with message score details.

Changes:

• Restricts verdicts to true_false scores while preserving missing and valueless score labels.
• Adds regression coverage for numeric and unknown scores, accessible labels, and human-score selection.

File	Description
frontend/​src/​components/​Chat/​ObjectiveHeader.tsx	Checks score type before assigning a verdict and displays raw nonboolean values.
frontend/​src/​components/​Chat/​ObjectiveHeader.test.tsx	Adds parameterized tests for nonboolean score display and radio selection.

_{Review effort: Balanced}