Datasets:
prompt stringlengths 2.98k 123k | completion stringlengths 153 9.37k | turn_id stringlengths 32 82 | traj_id stringlengths 30 79 | source stringclasses 8
values | language stringclasses 5
values | phase stringclasses 4
values | stratum stringlengths 16 59 | n_prefix_chars int64 2.89k 119k | n_prefix_tokens int64 739 30k | teacher_lift float64 -0.08 0.68 | modal_share float64 0.33 1 | n_pool_refs int64 1 90 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
<|im_start|>system
You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.
Your response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).
Include a THOUGHT section before your command where you explain your reasoning... | </think>
THOUGHT: Now I need to explore the django_migration_linter directory to find the sql_analyser directory and the base.py file mentioned in the PR description. This is where the issue is located.
Next bash command to run now:
```bash
find django_migration_linter -name "base.py" -type f
``` | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:1 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | early | 3YOURMIND__django-migration-linter|pr_258 | 8,969 | 2,433 | 0.023534 | 0.5 | 4 |
<|im_start|>system
You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.
Your response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).
Include a THOUGHT section before your command where you explain your reasoning... | </think>
THOUGHT: Great! The test script ran successfully and found no errors, which means our fix is working correctly. The CREATE INDEX statement with a NOT NULL condition in the WHERE clause is no longer being flagged as a NOT NULL constraint error.
Let's also test a case that should still be flagged as an error to... | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:10 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | mid | 3YOURMIND__django-migration-linter|pr_258 | 22,727 | 6,058 | 0.026211 | 1 | 13 |
<|im_start|>system
You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.
Your response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).
Include a THOUGHT section before your command where you explain your reasoning... | </think>
THOUGHT: Now let's run the extended test to verify both cases: that our fix works for CREATE INDEX statements with NOT NULL, and that we haven't broken the detection of actual NOT NULL constraint additions.
Next bash command to run now:
```bash
python test_fix_extended.py
``` | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:11 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | mid | 3YOURMIND__django-migration-linter|pr_258 | 24,293 | 6,465 | 0.000281 | 1 | 4 |
<|im_start|>system
You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.
Your response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).
Include a THOUGHT section before your command where you explain your reasoning... | </think>
THOUGHT: Let's run the comprehensive test to ensure our fix works correctly across various edge cases.
Next bash command to run now:
```bash
python test_comprehensive.py
``` | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:13 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | late | 3YOURMIND__django-migration-linter|pr_258 | 27,217 | 7,231 | 0.000051 | 1 | 1 |
<|im_start|>system
You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.
Your response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).
Include a THOUGHT section before your command where you explain your reasoning... | </think>
THOUGHT: <think>
Excellent! All tests pass, confirming our fix is working correctly. Now we need to prepare our changes for submission. Let's create a patch file with the changes we made to the source code.
Next bash command to run now:
```bash
git diff django_migration_linter/sql_analyser/base.py > patch.tx... | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:14 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | late | 3YOURMIND__django-migration-linter|pr_258 | 27,717 | 7,375 | 0.059363 | 1 | 5 |
<|im_start|>system
You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.
Your response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).
Include a THOUGHT section before your command where you explain your reasoning... | </think>
THOUGHT: Now I should verify that my change was applied correctly by checking the modified file. I'll also want to create a test to ensure the fix works as expected.
Next bash command to run now:
```bash
cat django_migration_linter/sql_analyser/base.py | grep -A 5 -B 5 "CREATE INDEX"
``` | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:3 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | early | 3YOURMIND__django-migration-linter|pr_258 | 17,667 | 4,579 | 0.043219 | 0.666667 | 6 |
<|im_start|>system
You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.
Your response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).
Include a THOUGHT section before your command where you explain your reasoning... | </think>
THOUGHT: Now let's run the test script to verify that our fix works correctly. The script should show no errors for the CREATE INDEX statement with NOT NULL condition after our fix.
Next bash command to run now:
```bash
python test_fix.py
``` | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:5 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | early | 3YOURMIND__django-migration-linter|pr_258 | 19,083 | 4,974 | 0.007268 | 1 | 4 |
<|im_start|>system
You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.
Your response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).
Include a THOUGHT section before your command where you explain your reasoning... | </think>
THOUGHT: <think>
It seems we need to install Django to run the test. Let's install it and then run our test script again.
Next bash command to run now:
```bash
pip install django && python test_fix.py
``` | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:6 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | mid | 3YOURMIND__django-migration-linter|pr_258 | 19,821 | 5,183 | 0.120756 | 0.6 | 10 |
"<|im_start|>system\nYou are a helpful assistant that can interact multiple times with a computer sh(...TRUNCATED) | "</think>\nTHOUGHT: Now that Django is installed, let's run our test script again to verify that our(...TRUNCATED) | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:7 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | mid | 3YOURMIND__django-migration-linter|pr_258 | 20,952 | 5,540 | 0.004232 | 0.75 | 4 |
"<|im_start|>system\nYou are a helpful assistant that can interact multiple times with a computer sh(...TRUNCATED) | "</think>\nTHOUGHT: It seems there are additional dependencies required. Let's install the package i(...TRUNCATED) | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731:8 | 3YOURMIND__django-migration-linter.c0a50bf7.pr_258__7ea2c731 | swe | python | mid | 3YOURMIND__django-migration-linter|pr_258 | 21,726 | 5,767 | 0.246885 | 0.666667 | 3 |
End of preview. Expand in Data Studio
quillon-186cf379
SFT corpus for Bittensor SN120 (Affine), Reason v4 (weight_version_key=7).
This is iamPi/tessera-77d11909
with one change: a fixed cue is appended to the end of every thought.
tessera: </think>\nTHOUGHT: {z}\n\n{y}
quillon: </think>\nTHOUGHT: {z}\n\nNext bash command to run now:\n\n{y}
The row set is identical — same 18,138 turns, same prompts, same thoughts,
same actions, same metadata columns. Removing the cue reconstructs tessera's
completion column byte-for-byte. The two are therefore a clean A/B: any
difference in downstream Reason is attributable to the cue and nothing else.
Why the cue sits there
The separator between the thought and the action at scoring time is hardcoded
as \n\n inside the validator's inject_prompt(). split_rollout() strips
trailing whitespace, so whatever a model emits, the parsed z_A ends at the
cue and the teacher forces THOUGHT: {z}…now:\n\n{y}. Keeping the \n\n in
the training target is the only construction where the string trained on is
byte-identical to the string the teacher scores.
Columns
prompt (rendered with Qwen/Qwen3.5-9B's chat template, ending inside an
open <think>), completion, plus turn_id, traj_id, source,
language, phase, stratum, n_prefix_chars, n_prefix_tokens,
teacher_lift, modal_share, n_pool_refs.
TRL prompt-completion format; use with SFTTrainer.
Known caveats
- 33 rows already exceeded the 1,792-token rollout budget
(
max_thought_tokens + max_action_tokens) in tessera; the cue adds 8 tokens and pushes 3 more over, for 36 of 18,138 (0.20%). Those targets are longer than a model can emit in one eval rollout. - The cue contributes the tokens
bash/command/runtoz, which tips the validator's fuzzy leakage heuristic on 14 rows (0.08%) whose command literally invokesbash(e.g.cd /app/... && bash setup_core.sh). Those rows fail the teacher-side B licence at eval. Zero rows in tessera trip it.
Provenance
Derived from SN120's public duel artifacts and turn corpus. Built with
scripts/retarget_thought_suffix.py, which re-parses every rewritten row with
the validator's own split_rollout() and aborts on a single mismatch.
- Downloads last month
- 37