diff --git a/bex/tag_preprocessor/analyze.py b/bex/tag_preprocessor/analyze.py
index a8be5cc..8180039 100644
--- a/bex/tag_preprocessor/analyze.py
+++ b/bex/tag_preprocessor/analyze.py
@@ -246,7 +246,7 @@ def _preprocess_files(file_paths):
     return sequences, seq_files
 
 
-def analyze_clusters(file_paths, extension, project_root="", min_coverage=DEFAULT_COVERAGE, prefer=None, kmax=2, N=3, include_kore=False, method='langsize', min_methods=5):
+def analyze_clusters(file_paths, extension, project_root="", min_coverage=DEFAULT_COVERAGE, prefer=None, kmax=2, N=3, include_kore=False, method='langsize', min_methods=3):
     """Run full pipeline: preprocess → frequency filter → ensemble infer.
 
     Returns:
@@ -283,7 +283,7 @@ def analyze_clusters(file_paths, extension, project_root="", min_coverage=DEFAUL
     return [("(all methods)", result, len(sequences), meta)]
 
 
-def _infer_group(label, group_seqs, group_files, project_root, min_coverage, prefer, kmax, N, include_kore=False, include_idregex=False, method='langsize', min_methods=5, crx_method='standard', min_structure=0.0):
+def _infer_group(label, group_seqs, group_files, project_root, min_coverage, prefer, kmax, N, include_kore=False, include_idregex=False, method='langsize', min_methods=3, crx_method='standard', min_structure=0.0):
     """Infer grammar for one package group. Module-level for ProcessPoolExecutor."""
     filtered = frequency_filter(group_seqs, min_coverage=min_coverage)
     imports = _extract_imports(group_files)
@@ -329,7 +329,7 @@ def _infer_group(label, group_seqs, group_files, project_root, min_coverage, pre
     return (label, result, len(filtered), meta)
 
 
-def analyze_by_package(file_paths, extension, project_root="", min_coverage=DEFAULT_COVERAGE, prefer=None, kmax=2, N=3, min_pkg_size=3, include_kore=False, include_idregex=False, method='langsize', min_methods=5, crx_method='standard', min_structure=0.0):
+def analyze_by_package(file_paths, extension, project_root="", min_coverage=DEFAULT_COVERAGE, prefer=None, kmax=2, N=3, min_pkg_size=3, include_kore=False, include_idregex=False, method='langsize', min_methods=3, crx_method='standard', min_structure=0.0):
     """Preprocess and group by package directory, infer per group.
 
     Groups methods by their file's relative directory path, merging
@@ -475,7 +475,7 @@ def analyze_directory(
     include_kore=False,
     include_idregex=False,
     method='langsize',
-    min_methods=5,
+    min_methods=3,
     crx_method='standard',
     min_structure=0.0,
 ):
@@ -679,8 +679,8 @@ def _parse_args(argv=None):
         help="Scoring method: langsize (default, Bex et al.) or mdl (fallback)",
     )
     parser.add_argument(
-        "--min-methods", type=int, default=5,
-        help="Minimum methods per group to infer grammar (default: 5). Groups with fewer are skipped.",
+        "--min-methods", type=int, default=3,
+        help="Minimum methods per group to infer grammar (default: 3). Groups with fewer are skipped.",
     )
     parser.add_argument(
         "--crx-method", choices=["standard", "refined"], default="standard",