Make strong incumbent trigger validation phase

src/aituner/harness.py

@@ -404,12 +404,11 @@ def _convergence_guard(
         and not infeasible_progress["plateau_detected"]
         and strong_incumbent["guard_active"]
     ):
-        reason = str(strong_incumbent["reason"])
+        reason = "strong_incumbent_requires_validation_probes"
     return {
         "should_stop_if_no_harness_can_justify_a_new_adjacent_probe": (
             should_stop
             or bool(infeasible_progress["stop_if_next_probe_repeats_family"])
-            or bool(strong_incumbent["guard_active"])
         ),
         "reason": reason,
         "infeasible_progress": infeasible_progress,
@@ -455,10 +454,14 @@ def _strong_incumbent_guard(
     if state.best_trial_id == latest.get("trial_id") and gain >= 1.8:
         return {
             "guard_active": True,
-            "reason": "incumbent_exceeds_baseline_by_1_8x_and_latest_trial_is_best",
+            "reason": "incumbent_exceeds_baseline_by_1_8x_and_latest_trial_is_best_enter_validation_phase",
             "baseline_trial_id": baseline.get("trial_id"),
             "baseline_request_rate_per_gpu": baseline_rate,
             "incumbent_gain_vs_baseline": gain,
+            "recommended_next_action": (
+                "validate_the_incumbent_with_adjacent_topology_or_same_topology_runtime_probes; "
+                "do_not_claim_global_optimum_from_the_first_large_gain"
+            ),
         }
     return {
         **default,
@@ -466,6 +469,7 @@ def _strong_incumbent_guard(
         "baseline_request_rate_per_gpu": baseline_rate,
         "incumbent_gain_vs_baseline": gain,
         "reason": "need_more_evidence_before_strong_incumbent_stop",
+        "recommended_next_action": "continue_harness_guided_search",
     }
 
 
@@ -600,7 +604,9 @@ def _proposal_rules() -> list[str]:
         "For decode_tpot bottlenecks, prefer legal TP/DP topology redistribution before runtime-only knobs, then tune max-num-seqs or max-num-batched-tokens only from observed SLO headroom/failures.",
         "Pick at most one primary knob family from knob_harnesses unless the history proves a coupled change is needed.",
         "Use adjacent legal values around the incumbent; avoid broad exploratory jumps.",
-        "When strong_incumbent.guard_active is true, do not propose runtime-only tweaks unless the relevant harness guard is positively satisfied by same-topology evidence.",
+        "When strong_incumbent.guard_active is true, treat the run as a validation phase, not as proof of optimum: probe only adjacent topology or same-topology runtime knobs that can falsify the incumbent.",
+        "Do not stop solely because a strong incumbent appeared; stop only after nearby validation probes are exhausted, blocked by infeasible evidence, or repeatedly remain near the incumbent.",
+        "During validation, prefer probes that answer whether the incumbent is locally optimal over probes that merely chase another large gain.",
         "Do not enable expert parallel for TTFT-prefill bottlenecks unless expert-parallel is the active harness and there is direct positive EP evidence.",
         "If infeasible_progress blocks the last primary knob family, do not continue that family; switch families with direct bottleneck evidence or set should_stop=true.",
         "If a proposed config is likely to reduce request_rate_per_gpu under the active guard, set should_stop=true instead of exploring.",

src/aituner/spec.py

@@ -63,6 +63,16 @@ def _coerce_str_list(value: Any, *, context: str) -> list[str]:
     return result
 
 
+def _coerce_text(value: Any, *, context: str) -> str:
+    if isinstance(value, str) and value.strip():
+        return value.strip()
+    if isinstance(value, Mapping) or isinstance(value, list):
+        text = json.dumps(value, ensure_ascii=False, sort_keys=True)
+        if text:
+            return text
+    raise SpecError(f"{context} must be a non-empty string.")
+
+
 def _coerce_int_list(value: Any, *, context: str) -> list[int]:
     if value is None:
         return []
@@ -674,8 +684,8 @@ class Proposal:
                 expected_effects, context="proposal.expected_effects"
             )
         return cls(
-            observation=_require_str(data.get("observation"), context="proposal.observation"),
-            diagnosis=_require_str(data.get("diagnosis"), context="proposal.diagnosis"),
+            observation=_coerce_text(data.get("observation"), context="proposal.observation"),
+            diagnosis=_coerce_text(data.get("diagnosis"), context="proposal.diagnosis"),
             config_patch=ConfigPatch.from_dict(
                 _require_mapping(data.get("config_patch"), context="proposal.config_patch")
             ),