From 62f2b894715a2daec2f7d4e8d81d0211bf30d27e Mon Sep 17 00:00:00 2001
From: RerankerGuo <121015044+RerankerGuo@users.noreply.github.com>
Date: Wed, 8 Jul 2026 14:11:01 +0800
Subject: [PATCH 1/2] feat(benchmark): add lme_one_question job wiring
 context_answer + answer_judge

---
 reme/config/jinli_lme.yaml | 35 +++++++++++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)

diff --git a/reme/config/jinli_lme.yaml b/reme/config/jinli_lme.yaml
index 6b6f76d6..aaf3c637 100644
--- a/reme/config/jinli_lme.yaml
+++ b/reme/config/jinli_lme.yaml
@@ -92,6 +92,41 @@ jobs:
     steps:
       - backend: python_execute_step
 
+  lme_one_question:
+    backend: base
+    description: "Run context_answer_step + answer_judge_step for a single LME question."
+    parameters:
+      type: object
+      properties:
+        query:
+          type: string
+          description: "the LME question being answered"
+        session_context:
+          type: string
+          description: "concatenated chat history for the relevant session"
+        current_date:
+          type: string
+          description: "today's date in YYYY-MM-DD, used to resolve relative phrases"
+        agent_answer:
+          type: string
+          description: "answer from the agent-under-test; can be the value of `context_answer` from a prior call"
+        golden_answer:
+          type: string
+          description: "ground-truth answer from the LME dataset"
+        question_type:
+          type: string
+          description: "LME category; one of temporal_reasoning / knowledge_update / single_session_preference / other"
+      required:
+        - query
+        - session_context
+        - current_date
+        - agent_answer
+        - golden_answer
+        - question_type
+    steps:
+      - backend: context_answer_step
+      - backend: answer_judge_step
+
 components:
   tokenizer:
     default:

From 3b1596972e90bb60355865232150b290c97e3e0f Mon Sep 17 00:00:00 2001
From: RerankerGuo <121015044+RerankerGuo@users.noreply.github.com>
Date: Wed, 8 Jul 2026 14:46:47 +0800
Subject: [PATCH 2/2] test(config): assert jinli_lme.yaml exposes
 lme_one_question with full required params

---
 tests/unit/test_config_parser.py | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/tests/unit/test_config_parser.py b/tests/unit/test_config_parser.py
index effe99b3..f1cdc0ff 100644
--- a/tests/unit/test_config_parser.py
+++ b/tests/unit/test_config_parser.py
@@ -86,3 +86,27 @@ def test_expand_env_vars_converts_expanded_scalar_types(monkeypatch):
         "url": "http://localhost:18080",
         "string_bool": "false",
     }
+
+
+def test_jinli_lme_config_registers_lme_one_question_job():
+    """``jinli_lme.yaml`` exposes a ``lme_one_question`` job that bundles the
+    LME ``context_answer`` and ``answer_judge`` steps and requires every
+    input the two steps consume."""
+    cfg = _load_config("jinli_lme.yaml")
+
+    job = cfg["jobs"]["lme_one_question"]
+    assert job["backend"] == "base"
+    assert [step["backend"] for step in job["steps"]] == [
+        "context_answer_step",
+        "answer_judge_step",
+    ]
+    assert job["parameters"]["required"] == [
+        "query",
+        "session_context",
+        "current_date",
+        "agent_answer",
+        "golden_answer",
+        "question_type",
+    ]
+    props = set(job["parameters"]["properties"])
+    assert props == set(job["parameters"]["required"])