diff --git a/src/mps_parser.c b/src/mps_parser.c
index 6119429..6b4bd2e 100644
--- a/src/mps_parser.c
+++ b/src/mps_parser.c
@@ -796,10 +796,26 @@ static int parse_columns_section(MpsParserState *state, char **tokens, int n_tok
     return 0;
 }
 
+// The set-name field of RHS, RANGES and BOUNDS lines is optional (Netlib and
+// CUTEst/SIF files often omit it). If the first data field already names a
+// known row (or column, for BOUNDS), there is no set name; otherwise the
+// first field is the set name and is skipped.
+static bool is_known_row(const MpsParserState *state, const char *name)
+{
+    if (state->objective_row_name && strcmp(name, state->objective_row_name) == 0)
+        return true;
+    return namemap_get(&state->row_map, name) != -1;
+}
+
+static int row_data_start(const MpsParserState *state, char **tokens, int n_tokens)
+{
+    return (n_tokens > 0 && is_known_row(state, tokens[0])) ? 0 : 1;
+}
+
 static int parse_rhs_section(MpsParserState *state, char **tokens, int n_tokens)
 {
 
-    for (int i = 1; i + 1 < n_tokens; i += 2)
+    for (int i = row_data_start(state, tokens, n_tokens); i + 1 < n_tokens; i += 2)
     {
         const char *row_name = tokens[i];
         double value = atof(tokens[i + 1]);
@@ -832,7 +848,7 @@ static int parse_rhs_section(MpsParserState *state, char **tokens, int n_tokens)
 static int parse_ranges_section(MpsParserState *state, char **tokens, int n_tokens)
 {
 
-    for (int i = 1; i + 1 < n_tokens; i += 2)
+    for (int i = row_data_start(state, tokens, n_tokens); i + 1 < n_tokens; i += 2)
     {
         const char *row_name = tokens[i];
         double range_val = atof(tokens[i + 1]);
@@ -870,13 +886,17 @@ static int parse_ranges_section(MpsParserState *state, char **tokens, int n_toke
 
 static int parse_bounds_section(MpsParserState *state, char **tokens, int n_tokens)
 {
-    if (n_tokens < 3)
+    if (n_tokens < 2)
         return 0;
 
     const char *bound_type = tokens[0];
 
-    const char *col_name = tokens[2];
-    double value = (n_tokens > 3) ? atof(tokens[3]) : 0.0;
+    // "type [set_name] column [value]"
+    int col_pos = (namemap_get(&state->col_map, tokens[1]) != -1) ? 1 : 2;
+    if (col_pos >= n_tokens)
+        return 0;
+    const char *col_name = tokens[col_pos];
+    double value = (col_pos + 1 < n_tokens) ? atof(tokens[col_pos + 1]) : 0.0;
 
     int col_idx = namemap_get(&state->col_map, col_name);
     if (col_idx == -1)
diff --git a/test/test_read_mps.py b/test/test_read_mps.py
index aa003cb..856e6af 100644
--- a/test/test_read_mps.py
+++ b/test/test_read_mps.py
@@ -51,6 +51,32 @@
     "OBJSENSE\n    MAXIMIZE\nROWS\n",
 )
 
+# MPS_MIN plus a RANGES entry, explicit bounds and an objective constant,
+# with the optional set-name field omitted in every RHS / RANGES / BOUNDS
+# line (as in Netlib / CUTEst files such as BLEND, SIERRA, GFRD-PNC, DFL001).
+MPS_NAMELESS = """NAME          TESTLP
+ROWS
+ N  COST
+ E  R1
+ L  R2
+ L  R3
+COLUMNS
+    X1        COST      1.0   R1        1.0
+    X1        R3        3.0
+    X2        COST      1.0   R1        2.0
+    X2        R2        1.0   R3        2.0
+RHS
+              R1        5.0   R2        2.0
+              R3        8.0   COST     -1.5
+RANGES
+              R3        6.0
+BOUNDS
+ UP           X1        4.0
+ LO           X2        0.5
+ MI           X1
+ENDATA
+"""
+
 
 @pytest.fixture
 def mps_min_file(tmp_path):
@@ -130,6 +156,26 @@ def test_read_objsense_maximize(mps_max_file):
     assert model.ModelSense == PDLP.MAXIMIZE
 
 
+def test_read_sections_without_set_name(tmp_path):
+    """
+    Entries without a set name must still reach the right rows/columns
+    (the reader used to drop the whole RHS section in this case).
+    """
+    path = tmp_path / "nameless.mps"
+    path.write_text(MPS_NAMELESS)
+    model = cupdlpx.read(str(path))
+    assert model.num_vars == 2
+    assert model.num_constrs == 3
+    assert np.allclose(model.c, [1.0, 1.0])
+    # RHS on the objective row is the negated objective constant
+    assert np.isclose(model.c0, 1.5)
+    # RHS and RANGES: R3 <= 8 with range 6 becomes 2 <= R3 <= 8
+    assert np.allclose(model.constr_lb, [5.0, -np.inf, 2.0])
+    assert np.allclose(model.constr_ub, [5.0, 2.0, 8.0])
+    assert np.allclose(model.lb, [-np.inf, 0.5])
+    assert np.allclose(model.ub, [4.0, np.inf])
+
+
 def test_read_missing_file_raises(tmp_path):
     """
     Reading a nonexistent file should raise FileNotFoundError.