Support full-range and incumbent-floor search modes

src/aituner/llm.py

@@ -342,8 +342,11 @@ def build_prompt(
         [
         "The primary cross-topology comparison metric is request_rate_per_gpu, not raw request_rate.",
         "The proposal should beat the incumbent on request_rate_per_gpu under the 95%+ SLO target.",
-        "The evaluator uses the best feasible sampling_u from the same tp_dp_product group when it exists.",
+        (
-        "If a tp_dp_product group has no history yet, the evaluator starts from the study's original search.low and runs a full binary search for that group.",
+            "The evaluator may use the same tp_dp_product incumbent as the search floor when search.inherit_incumbent_floor=true."
+            if study.search.inherit_incumbent_floor
+            else "The evaluator runs each proposal over the full configured search range so raw per-iteration performance is measured directly."
+        ),
         "Do not assume a configuration with fewer GPUs should inherit the global incumbent sampling_u.",
         (
             "Follow the active harness. Prefer stop over a weak exploratory proposal once a good incumbent has converged."

src/aituner/spec.py

@@ -511,6 +511,7 @@ class SamplingSearchSpec:
     tolerance: float
     max_probes: int
     sample_seed: int
+    inherit_incumbent_floor: bool = False
 
     @classmethod
     def from_dict(cls, data: Mapping[str, Any]) -> "SamplingSearchSpec":
@@ -524,6 +525,10 @@ class SamplingSearchSpec:
             sample_seed=_require_int(
                 data.get("sample_seed", 20260325), context="search.sample_seed"
             ),
+            inherit_incumbent_floor=_require_bool(
+                data.get("inherit_incumbent_floor", False),
+                context="search.inherit_incumbent_floor",
+            ),
         )
 
 

src/aituner/store.py

@@ -85,11 +85,21 @@ class StudyStore:
         trial_root = self.study_root(study.study_id) / "trials" / trial_id
         trial_root.mkdir(parents=True, exist_ok=True)
         parallel_size = _parallel_size_for_proposal(study=study, proposal=proposal)
+        search = study.search
+        if study.search.inherit_incumbent_floor:
+            search = replace(
+                study.search,
+                low=_derive_search_floor(
+                    study=study,
+                    state=state,
+                    parallel_size=parallel_size,
+                ),
+            )
         spec = TrialSpec(
             study_id=study.study_id,
             trial_id=trial_id,
             config_patch=proposal.config_patch,
-            search=study.search,
+            search=search,
             study_spec_path=str((self.study_root(study.study_id) / "study_spec.source").resolve()),
             artifact_dir=str(trial_root),
             probe_log_path=str(trial_root / "probe_history.json"),
@@ -301,3 +311,15 @@ def _request_rate_per_gpu(best_request_rate: Any, parallel_size: int | None) ->
         return None
     return float(best_request_rate) / float(parallel_size)
 
+
+def _derive_search_floor(*, study: StudySpec, state: StudyState, parallel_size: int) -> float:
+    low = study.search.low
+    high = study.search.high
+    group_incumbent = (state.best_by_parallel_size or {}).get(str(parallel_size))
+    if isinstance(group_incumbent, dict) and isinstance(
+        group_incumbent.get("best_sampling_u"), (int, float)
+    ):
+        candidate = float(group_incumbent["best_sampling_u"])
+    else:
+        candidate = low
+    return min(high, max(low, candidate))

tests/test_core_flow.py

@@ -2248,6 +2248,47 @@ class CoreFlowTests(unittest.TestCase):
             trial, _ = store.materialize_trial(study=study, state=state, proposal=proposal)
             self.assertEqual(trial.search.low, study.search.low)
 
+    def test_materialize_trial_can_use_incumbent_floor_when_enabled(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            tmp_path = Path(tmp)
+            study_path = _write_study_assets(tmp_path)
+            payload = json.loads(study_path.read_text(encoding="utf-8"))
+            payload["search"]["inherit_incumbent_floor"] = True
+            study_path.write_text(json.dumps(payload), encoding="utf-8")
+            study = load_study_spec(study_path)
+            store = StudyStore(tmp_path / ".aituner" / "studies")
+            store.init_study(spec_path=study_path, study=study)
+            state = StudyState(
+                study_id=study.study_id,
+                best_trial_id="trial-0001",
+                best_parallel_size=4,
+                best_sampling_u=0.375,
+                best_request_rate=3.0,
+                best_request_rate_per_gpu=0.75,
+                next_trial_index=2,
+                best_by_parallel_size={
+                    "4": {
+                        "trial_id": "trial-0001",
+                        "parallel_size": 4,
+                        "best_sampling_u": 0.375,
+                        "best_request_rate": 3.0,
+                        "best_request_rate_per_gpu": 0.75,
+                    }
+                },
+                trials=[],
+            )
+            proposal = Proposal.from_dict(
+                {
+                    "observation": "Obs",
+                    "diagnosis": "Diag",
+                    "config_patch": {"env_patch": {}, "flag_patch": {"tensor-parallel-size": 4}},
+                    "expected_effects": ["raise rate"],
+                }
+            )
+            trial, _ = store.materialize_trial(study=study, state=state, proposal=proposal)
+            self.assertEqual(trial.search.low, 0.375)
+            self.assertTrue(trial.search.inherit_incumbent_floor)
+
     def test_materialize_trial_resets_search_floor_for_new_parallel_group(self) -> None:
         with tempfile.TemporaryDirectory() as tmp:
             tmp_path = Path(tmp)