Use dedicated single-video prompt template

single_video_567/code/unified_eval.py

@@ -23,13 +23,13 @@ from concurrent.futures import ThreadPoolExecutor, as_completed
 from pathlib import Path
 
 
-VERSION = "runningbench-unified-v6"
+VERSION = "runningbench-unified-v7"
 DATASETS = {
     "single_698": "segments_60s_gemini_verification/final/segments_700_verified.jsonl",
     "single_567_human": "human_answer_repair_567/single_567_human_eval.jsonl",
     "mixed_1487": "qa_fix_gold_recovery_and_adjudication/data/runningbench_v2_kept.jsonl",
 }
-PROMPT = """Answer this multiple-choice question about egocentric walking/running footage.
+MULTI_VIDEO_PROMPT = """Answer this multiple-choice question about egocentric walking/running footage.
 The frames below are sampled in order from the clip(s) the question refers to; each clip is
 introduced by its label.
 
@@ -40,6 +40,18 @@ QUESTION: {question}
 OPTIONS:
 {options}
 
+Select exactly {n} option{plural}.
+Return ONLY JSON: {{"answer": [{example}]}}"""
+SINGLE_VIDEO_PROMPT = """Answer this multiple-choice question about egocentric walking/running footage.
+The frames below are sampled in chronological order from the single 60-second video associated with this question.
+
+Watch the video frames before deciding. Base your answer on visual evidence from this video. Do not answer from the wording of the options alone.
+
+QUESTION: {question}
+
+OPTIONS:
+{options}
+
 Select exactly {n} option{plural}.
 Return ONLY JSON: {{"answer": [{example}]}}"""
 MULTI_UNITS = {"cross_video", "cross_recording", "anonymous_multi_clip"}
@@ -316,11 +328,12 @@ class FloodgateClient:
                 "attempts": [], "error": None}
 
 
-def prompt_for(question, qid):
+def prompt_for(question, qid, dataset):
     shown, back = shuffled_view(question["options"], qid)
     n = len(question["answer"])
     letters = sorted(shown)
-    prompt = PROMPT.format(question=question["question"], options="\n".join(f"{letter}. {shown[letter]}" for letter in letters),
+    template = SINGLE_VIDEO_PROMPT if dataset.startswith("single_") else MULTI_VIDEO_PROMPT
+    prompt = template.format(question=question["question"], options="\n".join(f"{letter}. {shown[letter]}" for letter in letters),
                            n=n, plural="s" if n > 1 else "", example=", ".join(f'"{letter}"' for letter in letters[:n]))
     return prompt, letters, back
 
@@ -342,8 +355,9 @@ def run(args, dataset, root, work):
     elif args.skip:
         questions = questions[args.skip:]
     run_dir = work / "runs" / args.run_id
+    prompt_template = SINGLE_VIDEO_PROMPT if dataset.startswith("single_") else MULTI_VIDEO_PROMPT
     config = {"version": VERSION, "dataset": dataset, "corpus_sha256": file_hash(corpus_path),
-              "prompt_sha256": hashlib.sha256(PROMPT.encode()).hexdigest(), "model": args.model,
+              "prompt_sha256": hashlib.sha256(prompt_template.encode()).hexdigest(), "model": args.model,
               "backend": args.backend,
               "thinking": args.thinking, "max_tokens": args.max_tokens,
               "max_tokens_cap": args.max_tokens_cap, "frames_per_question": args.frames,
@@ -395,7 +409,7 @@ def run(args, dataset, root, work):
         qid = question_id(question, dataset)
         if isinstance(frames, Exception):
             return {"question_id": qid, "dataset": dataset, "error": f"media:{type(frames).__name__}: {frames}"}
-        prompt, letters, back = prompt_for(question, qid)
+        prompt, letters, back = prompt_for(question, qid, dataset)
         started = time.time()
         try:
             response = client.ask(prompt, frames)