Initial Commit

2026-09-15 15:24:03 -04:00
commit 4dc7086e7e
605 changed files with 101380 additions and 0 deletions
@@ -0,0 +1,42 @@
+# e11 — Bat and ball, altered total
+
+| | |
+|---|---|
+| **Difficulty** | easy |
+| **Domain** | arithmetic / cognitive-reflection trap |
+| **Probes** | whether the model retrieves the memorised answer to the $1.10 version or solves the one asked |
+| **Tools expected** | none needed; one `matlab_run` is fine |
+| **Trap** | The famous version totals $1.10 and the famous answer is $0.05. This one totals $1.30. |
+
+## Ground truth
+
+ball = **$0.15**, bat = $1.15 (check: 0.15 + 1.15 = 1.30; 1.15 − 0.15 = 1.00).
+
+## Scoring (10 points)
+
+### Correctness (5)
+- [ ] (5) $0.15
+
+### Calibration & honesty (2)
+- [ ] (1) Shows the check (both conditions satisfied)
+- [ ] (1) No web search for a two-line algebra problem
+
+### Tool use (2)
+- [ ] (2) No tool, or a single arithmetic check. 0 if it searched the web for "bat and ball".
+
+### Communication (1)
+- [ ] (1) Answer first
+
+### Automatic deductions
+- −4 answers $0.05 or $0.30
+- −1 intent-without-call
+- automatic fail
+
+## Grading
+
+One worksheet per round, alongside that round's transcript:
+
+- `v1/grade.md` · `v1/result.txt` — first round (baseline)
+- `v2/grade.md` · `v2/result.txt` — second round
+
+Copy the scoring skeleton above into a new round's `grade.md` to add one.