code backup: harness

harness/B/launch.py

@@ -26,6 +26,7 @@ if str(WORKSPACE_ROOT) not in sys.path:
 from harness.A import EXTENDED_MAX_NEW_TOKENS, MAX_NEW_TOKENS  # noqa: E402
 from harness.A import models as vlm_models  # noqa: E402
 from harness.A.launch import scenes  # noqa: E402
+from harness.B.prompts import PARAPHRASE_PRE_PROMPT  # noqa: E402
 from harness.B import (  # noqa: E402
     DEFAULT_DEPTH,
     DEFAULT_INPUT_SELECTION,
@@ -51,6 +52,7 @@ def _load_run_module():
 def _worker(
     tasks, results, model, spatial_code_format, input_selection, frame_count, depth, tracking,
     results_dir, gpu, cpu_threads, extended, reasoning_budget, force_budget,
+    serialization, context_line,
 ):
     if gpu is not None:
         os.environ["CUDA_VISIBLE_DEVICES"] = str(gpu)
@@ -85,6 +87,8 @@ def _worker(
                 extended=extended,
                 reasoning_budget=reasoning_budget,
                 force_budget=force_budget,
+                serialization=serialization,
+                context_line=context_line,
             )
             mean_score = (
                 sum(r["score"] for r in answered) / len(answered) if answered else None
@@ -100,6 +104,7 @@ def launch(
     model, spatial_code_format, input_selection, frame_count, selected,
     depth=DEFAULT_DEPTH, tracking=DEFAULT_TRACKING, results_dir=None, rebuild=False,
     extended=True, reasoning_budget=EXTENDED_MAX_NEW_TOKENS, force_budget=MAX_NEW_TOKENS,
+    serialization="json", context_line=None,
 ):
     """Answer every question for ``selected`` scenes, sharded across every visible GPU."""
     protocol = "extended" if extended else "base"
@@ -149,7 +154,7 @@ def launch(
             args=(
                 tasks, results, model, spatial_code_format, input_selection, frame_count,
                 depth, tracking, results_dir, gpu, cpu_threads, extended, reasoning_budget,
-                force_budget,
+                force_budget, serialization, context_line,
             ),
         )
         for gpu in assignments
@@ -200,6 +205,16 @@ def main():
         "--base-protocol", action="store_true",
         help="run harness.A's exact fixed 16-token protocol instead of the extended default",
     )
+    parser.add_argument(
+        "--serialization", default="json",
+        help="robustness arm only: 'yaml' renders the identical code dict as YAML "
+        "(pair with an explicit --results-dir)",
+    )
+    parser.add_argument(
+        "--paraphrase-context", action="store_true", dest="paraphrase_context",
+        help="robustness arm only: the pre-registered paraphrased context line "
+        "(pair with an explicit --results-dir)",
+    )
     parser.add_argument("--reasoning-budget", type=int, default=EXTENDED_MAX_NEW_TOKENS)
     parser.add_argument("--force-budget", type=int, default=MAX_NEW_TOKENS)
     args = parser.parse_args()
@@ -223,6 +238,8 @@ def main():
         depth=args.depth, tracking=args.tracking, results_dir=args.results_dir, rebuild=args.rebuild,
         extended=not args.base_protocol,
         reasoning_budget=args.reasoning_budget, force_budget=args.force_budget,
+        serialization=args.serialization,
+        context_line=PARAPHRASE_PRE_PROMPT if args.paraphrase_context else None,
     )
 
 

harness/B/prompts.py

@@ -37,18 +37,54 @@ CODE_DESCRIPTION = (
 )
 PRE_PROMPT = "This is a spatial code: " + CODE_DESCRIPTION
 
+# The robustness arm's alternates (analysis/preregistration.md): a semantically
+# equivalent paraphrase of the context line, and a YAML rendering of the identical
+# code dict. Each changes exactly one surface property, so a result that survives
+# both cannot be an artifact of the specific wording or serialization chosen.
+PARAPHRASE_PRE_PROMPT = (
+    "Below is a spatial code -- a structured description, in JSON, of a room and the "
+    "objects inside it. Every field is explained by the schema embedded in the code."
+)
+SERIALIZATIONS = ("json", "yaml")
+
+
+def render_code(spatial_code, serialization="json"):
+    """Serialize one spatial-code dict for prompting. "json" is the standard arm;
+    "yaml" renders the IDENTICAL dict (key order preserved, nothing added or
+    dropped) for the serialization-robustness arm."""
+    if serialization == "json":
+        return json.dumps(spatial_code, indent=1)
+    if serialization == "yaml":
+        import yaml
+
+        return yaml.safe_dump(spatial_code, sort_keys=False, width=100)
+    raise ValueError(
+        f"unknown serialization {serialization!r}; expected one of {SERIALIZATIONS}"
+    )
+
 
-def build_prompt(spatial_code, question_type, question, options=None):
+def build_prompt(
+    spatial_code, question_type, question, options=None,
+    serialization="json", context_line=None,
+):
     """Return the full text prompt: context line, the spatial code itself, the question,
-    and the same VSI-Bench post-prompt harness.A uses for the same question_type."""
-    code_text = json.dumps(spatial_code, indent=1)
+    and the same VSI-Bench post-prompt harness.A uses for the same question_type.
+    ``serialization`` and ``context_line`` exist ONLY for the robustness arm -- the
+    defaults reproduce the standard prompt byte-for-byte."""
+    code_text = render_code(spatial_code, serialization)
+    pre_prompt = PRE_PROMPT if context_line is None else context_line
+    if serialization == "yaml" and context_line is None:
+        # The context line must not claim JSON when the code is rendered as YAML --
+        # otherwise the serialization arm would carry a false description as a
+        # second, unintended factor.
+        pre_prompt = pre_prompt.replace("JSON", "YAML")
     if question_type in NA_QUESTION_TYPES:
-        return "\n".join([PRE_PROMPT, code_text, question, NA_POST_PROMPT])
+        return "\n".join([pre_prompt, code_text, question, NA_POST_PROMPT])
     if question_type in MCA_QUESTION_TYPES:
         if not options:
             raise ValueError(f"question_type {question_type!r} requires options")
         options_block = "Options:\n" + "\n".join(options)
-        return "\n".join([PRE_PROMPT, code_text, question, options_block, MCA_POST_PROMPT])
+        return "\n".join([pre_prompt, code_text, question, options_block, MCA_POST_PROMPT])
     raise ValueError(
         f"unknown question_type {question_type!r}; "
         f"expected one of {MCA_QUESTION_TYPES + NA_QUESTION_TYPES}"

harness/B/run.py

@@ -36,6 +36,7 @@ from harness.B import (  # noqa: E402
 )
 from harness.B import prompts as code_prompts  # noqa: E402
 from harness.B import spatial_codes  # noqa: E402
+from harness.B.prompts import PARAPHRASE_PRE_PROMPT, SERIALIZATIONS  # noqa: E402
 
 
 def results_dir_for(
@@ -146,6 +147,8 @@ def run(
     extended=True,
     reasoning_budget=EXTENDED_MAX_NEW_TOKENS,
     force_budget=MAX_NEW_TOKENS,
+    serialization="json",
+    context_line=None,
 ):
     """Answer every matching question with one model, given its scene's spatial code as
     text (no video frames). Each question's full record is written to its own JSON file
@@ -182,7 +185,8 @@ def run(
                 code_cache[scene_id] = {"code": code, "path": path}
             cached = code_cache[scene_id]
             prompt = code_prompts.build_prompt(
-                cached["code"], row["question_type"], row["question"], row.get("options")
+                cached["code"], row["question_type"], row["question"], row.get("options"),
+                serialization=serialization, context_line=context_line,
             )
             answer = (
                 adapter.answer_extended(
@@ -249,6 +253,16 @@ def main():
         "--no-write", action="store_true",
         help="skip writing per-question JSON files; print/score only",
     )
+    parser.add_argument(
+        "--serialization", default="json", choices=SERIALIZATIONS,
+        help="robustness arm only: render the identical code dict as YAML instead of "
+        "JSON (pair with an explicit --results-dir so the arm stays isolated)",
+    )
+    parser.add_argument(
+        "--paraphrase-context", action="store_true", dest="paraphrase_context",
+        help="robustness arm only: use the pre-registered paraphrased context line "
+        "(pair with an explicit --results-dir)",
+    )
     parser.add_argument(
         "--base-protocol", action="store_true",
         help="run harness.A's exact fixed 16-token protocol (plain answer()) instead of "
@@ -277,6 +291,8 @@ def main():
         results_dir=args.results_dir,
         write_results=not args.no_write,
         extended=not args.base_protocol,
+        serialization=args.serialization,
+        context_line=PARAPHRASE_PRE_PROMPT if args.paraphrase_context else None,
         reasoning_budget=args.reasoning_budget,
         force_budget=args.force_budget,
     )