LASR-Callum/2026-08-01-qwen36-difficult-advice-tulu-lora-40-60-empty-think-tags
Image-Text-to-Text • Updated • 12
Datasets:
Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code: DatasetGenerationError
Exception: CastError
Message: Couldn't cast
text: string
source: string
spans: list<item: list<item: int64>>
child 0, item: list<item: int64>
child 0, item: int64
row: int64
n_tokens: int64
n_supervised: int64
to
{'row': Value('int64'), 'source': Value('string'), 'n_tokens': Value('int64'), 'n_supervised': Value('int64'), 'spans': List(List(Value('int64')))}
because column names don't match
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1827, in _prepare_split_single
for key, table in generator:
^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
for item in generator(*args, **kwargs):
~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
text: string
source: string
spans: list<item: list<item: int64>>
child 0, item: list<item: int64>
child 0, item: int64
row: int64
n_tokens: int64
n_supervised: int64
to
{'row': Value('int64'), 'source': Value('string'), 'n_tokens': Value('int64'), 'n_supervised': Value('int64'), 'spans': List(List(Value('int64')))}
because column names don't match
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1880, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
row int64 | source string | n_tokens int64 | n_supervised int64 | spans list |
|---|---|---|---|---|
0 | difficult_advice | 990 | 831 | [
[
700,
4927
]
] |
1 | tulu3 | 378 | 77 | [
[
1395,
1850
]
] |
2 | difficult_advice | 1,088 | 959 | [
[
647,
5377
]
] |
3 | tulu3 | 1,111 | 1,016 | [
[
256,
2755
]
] |
4 | tulu3 | 249 | 185 | [
[
300,
1238
]
] |
5 | difficult_advice | 1,106 | 944 | [
[
762,
5480
]
] |
6 | tulu3 | 83 | 54 | [
[
159,
454
]
] |
7 | tulu3 | 247 | 107 | [
[
740,
1238
]
] |
8 | tulu3 | 62 | 37 | [
[
113,
266
]
] |
9 | difficult_advice | 977 | 838 | [
[
620,
4934
]
] |
10 | difficult_advice | 953 | 801 | [
[
779,
4731
]
] |
11 | tulu3 | 287 | 238 | [
[
80,
541
],
[
684,
1186
]
] |
12 | difficult_advice | 1,033 | 874 | [
[
826,
5487
]
] |
13 | tulu3 | 191 | 130 | [
[
223,
643
]
] |
14 | difficult_advice | 928 | 761 | [
[
695,
4441
]
] |
15 | tulu3 | 1,689 | 1,633 | [
[
89,
3254
],
[
3391,
6334
],
[
6460,
9641
]
] |
16 | tulu3 | 318 | 283 | [
[
124,
799
]
] |
17 | tulu3 | 149 | 111 | [
[
94,
425
],
[
518,
611
]
] |
18 | tulu3 | 371 | 333 | [
[
152,
1724
]
] |
19 | difficult_advice | 1,167 | 1,048 | [
[
613,
5872
]
] |
20 | difficult_advice | 966 | 834 | [
[
686,
5006
]
] |
21 | difficult_advice | 1,104 | 940 | [
[
732,
5596
]
] |
22 | tulu3 | 1,184 | 904 | [
[
1185,
3716
]
] |
23 | tulu3 | 1,806 | 1,424 | [
[
1388,
4213
]
] |
24 | tulu3 | 679 | 609 | [
[
207,
1880
]
] |
25 | difficult_advice | 930 | 797 | [
[
673,
4882
]
] |
26 | tulu3 | 1,251 | 988 | [
[
651,
4485
]
] |
27 | tulu3 | 877 | 844 | [
[
125,
3617
]
] |
28 | tulu3 | 217 | 198 | [
[
109,
1049
]
] |
29 | tulu3 | 378 | 287 | [
[
445,
1724
]
] |
30 | tulu3 | 1,115 | 874 | [
[
1156,
3480
]
] |
31 | difficult_advice | 1,133 | 962 | [
[
802,
5536
]
] |
32 | tulu3 | 1,759 | 1,400 | [
[
1473,
5350
]
] |
33 | difficult_advice | 1,125 | 947 | [
[
771,
5498
]
] |
34 | tulu3 | 295 | 233 | [
[
311,
1598
]
] |
35 | tulu3 | 27 | 6 | [
[
97,
134
]
] |
36 | difficult_advice | 1,071 | 906 | [
[
848,
5820
]
] |
37 | difficult_advice | 1,004 | 876 | [
[
625,
5185
]
] |
38 | tulu3 | 1,149 | 949 | [
[
944,
3236
]
] |
39 | tulu3 | 731 | 585 | [
[
734,
2853
]
] |
40 | difficult_advice | 1,094 | 916 | [
[
832,
5752
]
] |
41 | difficult_advice | 1,020 | 888 | [
[
638,
5322
]
] |
42 | difficult_advice | 992 | 826 | [
[
835,
5334
]
] |
43 | tulu3 | 853 | 441 | [
[
1317,
2703
]
] |
44 | tulu3 | 290 | 231 | [
[
127,
315
],
[
448,
1108
]
] |
45 | difficult_advice | 1,162 | 982 | [
[
934,
6012
]
] |
46 | tulu3 | 1,005 | 680 | [
[
1317,
3303
]
] |
47 | difficult_advice | 1,097 | 973 | [
[
601,
5520
]
] |
48 | difficult_advice | 1,257 | 1,118 | [
[
780,
6704
]
] |
49 | tulu3 | 394 | 282 | [
[
95,
625
],
[
809,
874
]
] |
50 | tulu3 | 263 | 203 | [
[
120,
562
],
[
705,
1034
]
] |
51 | tulu3 | 1,116 | 864 | [
[
962,
3901
]
] |
52 | difficult_advice | 1,219 | 1,057 | [
[
838,
6165
]
] |
53 | difficult_advice | 971 | 840 | [
[
661,
5073
]
] |
54 | tulu3 | 907 | 696 | [
[
604,
3357
]
] |
55 | difficult_advice | 1,084 | 902 | [
[
841,
5439
]
] |
56 | difficult_advice | 890 | 742 | [
[
670,
4507
]
] |
57 | tulu3 | 73 | 52 | [
[
112,
352
]
] |
58 | tulu3 | 989 | 807 | [
[
912,
3382
]
] |
59 | tulu3 | 85 | 53 | [
[
166,
436
]
] |
60 | difficult_advice | 972 | 849 | [
[
555,
4898
]
] |
61 | difficult_advice | 989 | 855 | [
[
632,
5013
]
] |
62 | difficult_advice | 1,154 | 980 | [
[
934,
6210
]
] |
63 | tulu3 | 393 | 270 | [
[
424,
1370
]
] |
64 | tulu3 | 203 | 93 | [
[
586,
1131
]
] |
65 | tulu3 | 1,836 | 1,438 | [
[
1668,
4941
]
] |
66 | tulu3 | 236 | 41 | [
[
1020,
1247
]
] |
67 | difficult_advice | 910 | 796 | [
[
567,
4720
]
] |
68 | tulu3 | 542 | 394 | [
[
625,
2110
]
] |
69 | tulu3 | 386 | 347 | [
[
176,
1606
]
] |
70 | difficult_advice | 1,034 | 870 | [
[
818,
5276
]
] |
71 | difficult_advice | 1,301 | 1,111 | [
[
911,
6634
]
] |
72 | difficult_advice | 960 | 789 | [
[
819,
4893
]
] |
73 | tulu3 | 209 | 190 | [
[
108,
1153
]
] |
74 | tulu3 | 233 | 127 | [
[
559,
1305
]
] |
75 | tulu3 | 261 | 77 | [
[
1021,
1474
]
] |
76 | tulu3 | 254 | 228 | [
[
138,
1261
]
] |
77 | difficult_advice | 1,020 | 898 | [
[
608,
5349
]
] |
78 | tulu3 | 245 | 185 | [
[
328,
1163
]
] |
79 | difficult_advice | 1,144 | 984 | [
[
741,
5858
]
] |
80 | difficult_advice | 1,055 | 928 | [
[
656,
5492
]
] |
81 | difficult_advice | 1,050 | 911 | [
[
756,
5740
]
] |
82 | tulu3 | 117 | 91 | [
[
131,
622
]
] |
83 | tulu3 | 597 | 555 | [
[
139,
1540
]
] |
84 | difficult_advice | 917 | 762 | [
[
780,
4801
]
] |
85 | tulu3 | 218 | 121 | [
[
515,
1125
]
] |
86 | tulu3 | 480 | 392 | [
[
269,
1277
]
] |
87 | tulu3 | 81 | 18 | [
[
258,
343
]
] |
88 | difficult_advice | 1,157 | 1,036 | [
[
636,
5896
]
] |
89 | tulu3 | 325 | 252 | [
[
324,
1312
]
] |
90 | tulu3 | 808 | 711 | [
[
202,
1783
]
] |
91 | tulu3 | 1,433 | 440 | [
[
4410,
6730
]
] |
92 | tulu3 | 966 | 116 | [
[
4214,
4823
]
] |
93 | difficult_advice | 959 | 825 | [
[
700,
5129
]
] |
94 | tulu3 | 269 | 194 | [
[
170,
692
]
] |
95 | tulu3 | 1,578 | 1,227 | [
[
1699,
5511
]
] |
96 | tulu3 | 531 | 476 | [
[
171,
1329
]
] |
97 | tulu3 | 820 | 793 | [
[
130,
3128
]
] |
98 | tulu3 | 198 | 161 | [
[
154,
851
]
] |
99 | tulu3 | 160 | 105 | [
[
266,
633
]
] |
End of preview.
Qwen3.6-27B SFT mixture — 40_60_empty_think_tags
40% difficult-advice / 60% TULU3 replay, with Qwen3.6's empty think marker added to the
replay rows and excluded from the loss. Training data for
qwen3.6-27b-difficult-advice-tulu-lora-40_60_empty_think_tags.
Derived from qwen3.6-27b-sft-mixture-40-60_assistant_loss_only — same rows,
same 580/1402 split, same seed. Only the markers differ. This file: md5 a09b35d6cd04c65616e7f0927d209bfe.
The marker
Every TULU3 replay row carries <think>\n\n</think>\n\n on its final assistant turn --
Qwen3.6's non-thinking marker, placed exactly where apply_chat_template puts it (the template
emits it only on final turns; the insertion is asserted to reproduce the template byte-for-byte
before any data is touched).
Those marker tokens are masked out of the loss. The model is conditioned on the marker --
which is how Qwen3.6 injects it as a prefill in non-thinking mode -- but never trained to emit
it, since learning to emit an empty think block is the documented reasoning-collapse pattern.
Difficult-advice rows are untouched and keep their real <think> traces fully supervised.
<|im_start|> MASKED
assistant MASKED
<think> MASKED <- marker: context, not a target
</think> MASKED
To LOSS <- supervision starts at the answer
| Source | Rows | Tokens | Marker | Supervised |
|---|---|---|---|---|
| difficult-advice | 580 | 597,013 | 0 | 85.48% |
| TULU3 replay | 1,402 | 901,954 | 1,402 | 77.5% |
| Total | 1,982 | 1,498,967 | 1,402 | 80.68% |
Supervision is otherwise assistant-tokens-only: everything outside an assistant turn is -100.
A supervised span ends after the closing <|im_end|>, which the model must produce to stop.
Files
| File | What it is |
|---|---|
mixture.jsonl |
the training input: text (pre-rendered) + source |
assistant_spans.jsonl |
per row, the character spans that carried loss (marker already excluded) |
stats.json |
the table above, machine-readable |
Re-rendering from messages will not reproduce these strings.
Provenance
src/experiments/add_empty_think.py applied to the published no-marker mixture above, itself built
by build_mixture.py (seed 0) over
allenai/tulu-3-sft-mixture and
matboz/difficult-advice-qwen3.
- Downloads last month
- 174
Models trained or fine-tuned on LASR-Callum/2026-08-01-qwen36-sft-mixture-40-60-empty-think-tags