diff --git a/serve/pd_server.py b/serve/pd_server.py
index d94672e..94bd4cb 100644
--- a/serve/pd_server.py
+++ b/serve/pd_server.py
@@ -71,6 +71,8 @@ def parse_args() -> argparse.Namespace:
     p.add_argument("--hicache-state-slots", type=int, default=4096)
     p.add_argument("--routing", default="affinity", choices=["affinity", "least-loaded"])
     p.add_argument("--prefill-any-interval", type=int, default=2)
+    p.add_argument("--eplb-interval", type=int, default=0, help="see openai_server.py; 0 (default) disables EPLB")
+    p.add_argument("--eplb-after-prefill", type=int, default=0, choices=[0, 1], help="see openai_server.py")
     return p.parse_args()
 
 
@@ -103,7 +105,10 @@ def worker_main(args: argparse.Namespace, dp_rank: int, in_q: mp.Queue, out_q: m
         os.environ["DEEPINFER_PREFILL_STORE_EXTRA_STATE"] = "1"
     os.environ.setdefault("DEEPINFER_PREFILL_STORE_EXTRA_SIZE", "512")  # identical state layout on PE and CE
     os.environ.setdefault("DEEPINFER_PREFILL_ANY_ATTEMPT_INTERVAL", str(args.prefill_any_interval))
-    setup_common_env(model_name=args.model, sampler_force_topk=1024, set_override_experts=True, enable_eplb=True)
+    setup_common_env(model_name=args.model, sampler_force_topk=1024, set_override_experts=True, enable_eplb=False)
+    if args.eplb_interval > 0:
+        os.environ["DEEPINFER_EXPERT_REARRANGEMENT_STEP_INTERVAL"] = str(args.eplb_interval)
+        os.environ["DEEPINFER_REARRANGE_AFTER_PREFILL_IN_NORMAL_ENGINE"] = str(args.eplb_after_prefill)
 
     engine_type = EngineType.PROMPT if args.role == "pe" else EngineType.COMPLETION
     llm = create_llm(