free-recall explain tool: the quizzing-mechanic switch (owner 2026-09-27)

Sycophancy mitigation as a structural commitment: the model writes its
full correctAnswer (plus explanation, in that order) BEFORE the learner
sees the question; the commitment rides the persisted call args, is
revealed on submit, and grades his free text in the model's next reply —
the client never grades free recall.

- tools.explain setting (segmented 'Quizzing — Multiple choice | Free
  recall' in Agent settings' Tools group): free arms explain and drops
  quiz for BOTH agents (prober variant carries no nodeId, same pattern
  as the probe quiz); mechanics keys never pass filterEnabledTools —
  quiz stays unwritable (the 2026-09-19 wedge ruling stands)
- doctrine lives pi-natively in the tool descriptions (owner ruling —
  no conditional system-prompt chunks): explain carries the commitment
  rule; the MC option-building craft moved out of the prompt into the
  quiz description; shared/20-question-tools.md slims to the
  quiz-agnostic core
- display seams: parseExplainInput (LaTeX-debris restoration included),
  dock text-mode with an 'I don't know' row (kbd row 0), AnsweredCard
  reveal (The answer / Why, no verdict chips), [Explain answer] format
- phases/plan seams recognize both quizzing tools' nodeId tags; explain
  answers deliberately feed no ledger evidence (no client grading to
  key on — a future ruling can parse the model's grade)

1049/1049 tests, typecheck/lint/build green.
This commit is contained in:
amosblomqvist
2026-09-27 08:55:13 +03:00
parent 07d2c84461
commit 2f39766528
25 changed files with 784 additions and 68 deletions
+40
View File
@@ -16,6 +16,7 @@ import {
parseAskInput,
parseProbeDoneInput,
parseQuizInput,
parseExplainInput,
} from "../src/core/tool-parts.ts";
import { quizInputSchema, probeQuizInputSchema } from "../src/core/question-core.ts";
import { probeDoneInputSchema } from "../src/core/contracts/probe.ts";
@@ -266,3 +267,42 @@ describe("display-math newline rule (Option A, owner ruling 2026-09-27)", () =>
assert.equal(parsed.question, "Is \\(p \\neq q\\) valid?");
});
});
// ─── parseExplainInput (free-recall display parse, owner 2026-09-27) ───────
describe("parseExplainInput: the dock's display seam for explain calls", () => {
const good = {
question: "Explain why Mercury is hard to observe.",
explanation: "It never strays far from the Sun in our sky.",
correctAnswer: "Its small angular separation from the Sun.",
nodeId: "planet-order", // the teacher's field — STRIPPED by Clean, like quiz
details: "Think about where it appears at dusk.",
};
it("parses a valid call and strips the teacher-only nodeId", () => {
const parsed = parseExplainInput(good)!;
assert.equal(parsed.question, "Explain why Mercury is hard to observe.");
assert.equal(parsed.correctAnswer, "Its small angular separation from the Sun.");
assert.equal(parsed.details, "Think about where it appears at dusk.");
assert.ok(!("nodeId" in parsed), "the teacher-only key is cleaned away");
});
it("an empty commitment never arms the dock", () => {
assert.equal(parseExplainInput({ ...good, correctAnswer: " " }), undefined);
});
it("restores LaTeX escape debris inside math spans (the pi-ai streaming wobble)", () => {
// \f of \forall was eaten by the lenient args parse: U+000C + "orall".
const dirty = "Because \( orall x\) holds.";
const parsed = parseExplainInput({ ...good, correctAnswer: dirty })!;
assert.equal(parsed.correctAnswer, "Because \(\forall x\) holds.");
});
it("wrong-shape input never arms the dock (numeric strings DO coerce — the seam's leniency)", () => {
assert.equal(parseExplainInput("just a string"), undefined);
assert.equal(parseExplainInput({ explanation: "e", correctAnswer: "a" }), undefined, "missing question");
assert.equal(parseExplainInput({ question: "q", explanation: "e", correctAnswer: { nested: true } }), undefined);
// 42 → "42": the same numeric coercion the quiz seam pins.
assert.equal(parseExplainInput({ question: "q", explanation: "e", correctAnswer: 42 })?.correctAnswer, "42");
});
});