Accumulate findings across attempts

Background

On a retry the Orchestrator re-runs only the flagged gap sub-questions, then merges the new results into a running all_findings dict — keeping the answers that were already good and filling the gaps.

for i, sq in enumerate(sub_questions):
    all_findings[sq] = results[i]

Problem statement

Implement merge_findings(all_findings, sub_questions, results) updating all_findings in place and returning it.

Input

• all_findings — the running dict {sub_question: finding} (mutate in place).
• sub_questions — the sub-questions just executed.
• results — their findings, aligned by index.

Output

Returns all_findings after writing each sub_questions[i] → results[i].

Examples

Example 1 — fill a gap

Input:  all_findings = {"benchmark": "CodeContests"},
        sub_questions = ["impact"], results = ["cited 1000+ times"]
Output: {"benchmark": "CodeContests", "impact": "cited 1000+ times"}

Constraints

• Write each sub_questions[i] → results[i] into all_findings.
• Existing keys not in sub_questions are kept; matching keys are overwritten.
• Mutate and return the same dict.

Notes

• Accumulating (rather than discarding) is what makes a retry strictly better — attempt 2 sees everything attempt 1 found, plus the newly-filled gap.