Datasets:
HelenPeng
/
ExeWorldBench

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
596
* if it is not the final component, whether it produces a meaningful output capable of advancing the chain.
* Necessary passive auxiliary structures are allowed, but auxiliary structures that artificially reduce the difficulty of achieving the intended mechanical behavior should be penalized. For example, adding side walls to a ramp solely to prevent the ball from falling off simplifies the required mechanical construction.
Different mechanical implementations are acceptable as long as the component preserves its definition and core mechanical function.
### Domino Chain
A Domino Chain should exhibit genuine sequential propagation:
Incoming Object → First Domino → Second Domino → ... → Final Domino
The incoming object should primarily trigger the first domino. Each subsequent domino should primarily fall because of the preceding domino.
Penalize direct triggering of multiple dominoes, group/simultaneous falling, independent downstream toppling, or final states inconsistent with sequential propagation.
Do not judge the Domino Chain only by whether all dominoes eventually fall.
---
## B. CAUSAL CONNECTIVITY
For every `A → B` segment, determine whether A genuinely and mechanically triggers B.
A valid transition should show:
A produces mechanical output
→ the output physically reaches and acts on B
→ B responds because of that interaction.
Check:
* whether A's output can physically reach B;
* whether the contact, collision, release, catching, or transmission is visible and plausible;
* whether B responds after and because of A's action;
* whether positions, directions, distances, and motion magnitudes support the interaction;
* whether the interaction avoids unexplained remote triggering, teleportation, interpenetration, premature triggering, or implausible energy transfer.
Temporal order alone is not sufficient evidence of causality.
As an auxiliary test, ask:
**Would B still execute in essentially the same way without A's mechanical action?**
If yes, A may have been bypassed and the segment score should be reduced.
The complete system should form one continuous chain in which all required components and balls genuinely participate. Required components must not be decorative, isolated, independently activated, or bypassed.
---
## C. CAUSAL CHAIN BREAK RULE
If the causal chain fails at a segment:
1. Score the failing segment according to the degree of successful interaction and target-component execution actually observed.
2. If the failure prevents mechanical propagation from reaching downstream components, **all subsequent segments must receive 0**.
3. Later independent motion of downstream components cannot restore their scores.
For example:
C1 → C2 → C3 → C4
If C1 and C2 execute correctly but C2 fails to successfully trigger C3:
* `initial → C1`: score normally
* `C1 → C2`: score normally
* `C2 → C3`: reduced score according to observed partial success
* `C3 → C4`: 0
Only components reached through continuous mechanical propagation from `initial` are eligible for non-zero downstream segment scores.
---
## D. SEGMENT SCORING
For `initial → A`, evaluate A's execution.
For `A → B`, jointly evaluate A→B causal interaction and B's execution.
* **9–10:** Correct, complete, and mechanically plausible interaction/execution with negligible issues.
* **7–8:** Overall correct with a few non-critical issues.
* **5–6:** Effective propagation and execution occur, but with significant mechanical, causal, or completeness issues.
* **3–4:** Only partial propagation or execution succeeds; major problems are present.
* **1–2:** Minimal valid propagation or execution is present; the segment largely fails.
* **0:** No meaningful valid propagation or target-component execution occurs, or the segment is unreachable because the causal chain has already broken.
==================================================
GENERAL PRINCIPLES
==================================================
* There is no single correct mechanical implementation.
* Every required component must be instantiated and genuinely participate.
* Every required ball must genuinely participate in the overall causal chain.
* Different balls may take over propagation only through visible and credible physical interaction.
* Passive auxiliary structures are allowed, but additional functional components must not replace or bypass required components.
* Do not infer causality from temporal order alone.
* Do not ignore intermediate failures merely because the final state appears correct.
* Once the continuous causal chain breaks, later independent motion cannot restore downstream segment scores.
==================================================
OUTPUT