Dataset · Text generation
nemotron-student-fail-v41-clean-thinking
by Zhiyuan Hu zhiyuanhucs/nemotron-student-fail-v41-clean-thinking
DeepSeek-V4.1 reward-1 trajectories for tasks on which the Nemotron student did not obtain reward 1. This release was rebuilt from the complete reward-1 audit under v56-shell-quote-aware-package-probes+v57-target-source-recheck.
Dataset Card
DeepSeek-V4.1 reward-1 trajectories for tasks on which the Nemotron student did not obtain reward 1. This release was rebuilt from the complete reward-1 audit under v56-shell-quote-aware-package-probes+v57-target-source-recheck. 每个目录内的文件名都直接说明:是一题一条还是全部 trials、是否为 audit、thinking/action 是否干净。 本版重新检查了每条候选轨迹的实际 tool action。查询、下载或读取当前任务项目的已发布源码时, 即使 verifier 给出 reward=1,也不会进入上述训练文件。审计隔离清单位于 data/03allreward1trials/targetpublishedsourcelookupquarantineaudit.jsonl.gz 167 条 trial)。 data/04reward1successfulbuthackingfailed/ 也单独排除了 SWE-bench Pro Verified is exported under a separate benchmark path: The Pro collector continuously rerolls unresolved tasks. The current export has 138 verifier-correct…
Excerpt from the card by Zhiyuan Hu, licensed other.
Details
- Repository
- zhiyuanhucs/nemotron-student-fail-v41-clean-thinking
- Publisher
- Zhiyuan Hu
- Task category
- Text generation
- Tags
- Not stated by the source
- Size category
- Not stated by the source
- Languages
- en
- Revision
- 7d4038616e7280af7c08e3d281191250a8354c87
- Last updated
- 2026-09-29
Files
88 files, 470.5 MB in total.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| blocked_attempt_control_manifest.jsonl | Data | 30.2 KB | — |
| data/01_clean_thinking_and_actions/all_successful_trials.jsonl.gz | Data | 1.1 MB | 5a95cc2b4614 |
| data/01_clean_thinking_and_actions/one_trajectory_per_task.jsonl.gz | Data | 893.5 KB | 321aede00557 |
| data/02_clean_actions_dirty_thinking_removed/all_successful_trials.jsonl.gz | Data | 37.7 MB | 1eab7ca8f445 |
| data/02_clean_actions_dirty_thinking_removed/one_trajectory_per_task.jsonl.gz | Data | 13.6 MB | dc6ecb3e9a23 |
| data/03_all_reward1_trials/all_trials__audit_index.jsonl.gz | Data | 47.5 KB | 514af1c17f95 |
| data/03_all_reward1_trials/clean_actions_dirty_thinking_removed__all_trials.jsonl.gz | Data | 37.7 MB | 1eab7ca8f445 |
| data/03_all_reward1_trials/clean_thinking_and_actions__all_trials.jsonl.gz | Data | 1.1 MB | 5a95cc2b4614 |
| data/03_all_reward1_trials/raw_original_trajectories_before_thinking_redaction__action_clean_reward1.jsonl.gz | Data | 38.4 MB | baec4c8defa1 |
| data/03_all_reward1_trials/rejected_trials__audit_only.jsonl.gz | Data | 33.5 KB | 1fba9247b43f |
| data/03_all_reward1_trials/target_published_source_lookup__quarantine_audit.jsonl.gz | Data | 44.9 KB | 8b4a80ab33ee |
| data/04_reward1_successful_but_hacking_failed/all_guard_rejected_dirty_generations.jsonl.gz | Data | 1.3 MB | 93e548f73d3a |
| data/04_reward1_successful_but_hacking_failed/all_successful_trials.jsonl.gz | Data | 29.6 MB | d07ed37086c0 |
| data/04_reward1_successful_but_hacking_failed/all_trials__audit_index.jsonl.gz | Data | 15.9 KB | 845d70908852 |
| data/04_reward1_successful_but_hacking_failed/one_trajectory_per_task.jsonl.gz | Data | 14.1 MB | f93398e6f472 |
| data/04_reward1_successful_but_hacking_failed/strong_proof_guard_blocked_before_execution/all_successful_trials.jsonl.gz | Data | 14.2 MB | bb6f27e4ddc8 |
| data/04_reward1_successful_but_hacking_failed/strong_proof_guard_blocked_before_execution/one_trajectory_per_task.jsonl.gz | Data | 5.5 MB | 17afc41798a7 |
| data/04_reward1_successful_but_hacking_failed/target_published_source_lookup__quarantine_audit.jsonl.gz | Data | 7.7 KB | 856972f20c53 |
| data/04_reward1_successful_but_hacking_failed/weak_proof_no_successful_answer_access_detected/all_successful_trials.jsonl.gz | Data | 15.4 MB | b6a75fe46db3 |
| data/04_reward1_successful_but_hacking_failed/weak_proof_no_successful_answer_access_detected/one_trajectory_per_task.jsonl.gz | Data | 8.5 MB | 3f42fffbede7 |
| data/05_offline_no_guard_experiment/deepseek_v4_1/all_reward1_without_successful_hack.jsonl.gz | Data | 53 B | 27fe8834bfd5 |
| data/05_offline_no_guard_experiment/deepseek_v4_1/all_scored_tasks__audit.jsonl.gz | Data | 5.8 KB | dbb5606333b5 |
| data/05_offline_no_guard_experiment/deepseek_v4_1/attempted_test_changes_but_external_verifier_still_passed.jsonl.gz | Data | 62 B | 80937dea99ce |
| data/05_offline_no_guard_experiment/deepseek_v4_1/clean_thinking_and_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1/dirty_actions_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1/dirty_thinking_and_dirty_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1/dirty_thinking_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1/legacy_public_egress_untrusted_reward1.jsonl.gz | Data | 4.9 MB | 631e76807426 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_full_pool/all_reward1_without_successful_hack.jsonl.gz | Data | 15.0 MB | 1dad8ed14165 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_full_pool/all_scored_tasks__audit.jsonl.gz | Data | 78.2 KB | 1cc3747dda6d |
| data/05_offline_no_guard_experiment/deepseek_v4_1_full_pool/attempted_test_changes_but_external_verifier_still_passed.jsonl.gz | Data | 2.2 MB | acb7cb0d313e |
| data/05_offline_no_guard_experiment/deepseek_v4_1_full_pool/clean_thinking_and_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_full_pool/dirty_actions_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_full_pool/dirty_thinking_and_dirty_actions__reward1.jsonl.gz | Data | 11.1 MB | f3e6ba40344d |
| data/05_offline_no_guard_experiment/deepseek_v4_1_full_pool/dirty_thinking_only__reward1.jsonl.gz | Data | 3.8 MB | 759787ecf43a |
| data/05_offline_no_guard_experiment/deepseek_v4_1_full_pool/legacy_public_egress_untrusted_reward1.jsonl.gz | Data | 7.4 MB | 55ad09c8b573 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_low_reasoning_pilot/all_reward1_without_successful_hack.jsonl.gz | Data | 53 B | 27fe8834bfd5 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_low_reasoning_pilot/all_scored_tasks__audit.jsonl.gz | Data | 280 B | 66ee847e8eb8 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_low_reasoning_pilot/clean_thinking_and_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_low_reasoning_pilot/dirty_actions_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_low_reasoning_pilot/dirty_thinking_and_dirty_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_low_reasoning_pilot/dirty_thinking_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_prompt_only_pilot/all_reward1_without_successful_hack.jsonl.gz | Data | 91.5 KB | 832c56e62d63 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_prompt_only_pilot/all_scored_tasks__audit.jsonl.gz | Data | 472 B | 6c26c10f061d |
| data/05_offline_no_guard_experiment/deepseek_v4_1_prompt_only_pilot/clean_thinking_and_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_prompt_only_pilot/dirty_actions_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_prompt_only_pilot/dirty_thinking_and_dirty_actions__reward1.jsonl.gz | Data | 91.5 KB | b31c1a79caed |
| data/05_offline_no_guard_experiment/deepseek_v4_1_prompt_only_pilot/dirty_thinking_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/deepseek_v4_1_vs_nemotron_3_ultra/all_paired_tasks__audit.jsonl.gz | Data | 742 B | cf5c058d8d6c |
| data/05_offline_no_guard_experiment/deepseek_v4_1_vs_nemotron_3_ultra/v4_1_success_nemotron_failed.jsonl.gz | Data | 57 B | a47124ddadc5 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra/all_reward1_without_successful_hack.jsonl.gz | Data | 53 B | 27fe8834bfd5 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra/all_scored_tasks__audit.jsonl.gz | Data | 5.9 KB | ba27b8283677 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra/attempted_test_changes_but_external_verifier_still_passed.jsonl.gz | Data | 62 B | 80937dea99ce |
| data/05_offline_no_guard_experiment/nemotron_3_ultra/clean_thinking_and_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra/dirty_actions_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra/dirty_thinking_and_dirty_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra/dirty_thinking_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra/legacy_public_egress_untrusted_reward1.jsonl.gz | Data | 1.0 MB | ff71ecf49df8 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra_step256_retry/all_reward1_without_successful_hack.jsonl.gz | Data | 53 B | 27fe8834bfd5 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra_step256_retry/all_scored_tasks__audit.jsonl.gz | Data | 277 B | a309005b0cef |
| data/05_offline_no_guard_experiment/nemotron_3_ultra_step256_retry/attempted_test_changes_but_external_verifier_still_passed.jsonl.gz | Data | 62 B | 80937dea99ce |
| data/05_offline_no_guard_experiment/nemotron_3_ultra_step256_retry/clean_thinking_and_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra_step256_retry/dirty_actions_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra_step256_retry/dirty_thinking_and_dirty_actions__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra_step256_retry/dirty_thinking_only__reward1.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/05_offline_no_guard_experiment/nemotron_3_ultra_step256_retry/legacy_public_egress_untrusted_reward1.jsonl.gz | Data | 65 B | a064e420ef57 |
| data/06_swebench_pro_verified/all_reward1_trials__audit_index.jsonl.gz | Data | 202.2 KB | 67c36d4097ef |
| data/06_swebench_pro_verified/clean_actions_dirty_thinking_removed__all_trials.jsonl.gz | Data | 34.4 MB | a2b575529960 |
| data/06_swebench_pro_verified/clean_actions_dirty_thinking_removed__one_per_task.jsonl.gz | Data | 30.7 MB | 31187b1a1098 |
| data/06_swebench_pro_verified/clean_thinking_and_actions__all_trials.jsonl.gz | Data | 40 B | eec3ad3a60e1 |
| data/06_swebench_pro_verified/clean_thinking_and_actions__one_per_task.jsonl.gz | Data | 32 B | 9022aad581c6 |
| data/07_official_swebench_pro_verified_baseline/deepseek_v4_1/all_valid_once_per_task.jsonl.gz | Data | 69.8 MB | 531617460705 |
| data/07_official_swebench_pro_verified_baseline/deepseek_v4_1/infrastructure_failures_excluded.jsonl.gz | Data | 5.6 KB | 653094249f4b |
| data/07_official_swebench_pro_verified_baseline/deepseek_v4_1/resolved_once_per_task.jsonl.gz | Data | 27.2 MB | e38e1fdfcf06 |
| data/07_official_swebench_pro_verified_baseline/nemotron_3_ultra/all_valid_once_per_task.jsonl.gz | Data | 30.5 MB | 7a7bf425c702 |
| data/07_official_swebench_pro_verified_baseline/nemotron_3_ultra/infrastructure_failures_excluded.jsonl.gz | Data | 2.5 KB | dd3cc2a3ddcd |
| data/07_official_swebench_pro_verified_baseline/nemotron_3_ultra/resolved_once_per_task.jsonl.gz | Data | 11.8 MB | c10ed24426fc |
| data/08_source_link_sanitized_recovery/clean_actions_dirty_thinking_removed__one_per_task.jsonl.gz | Data | 325.5 KB | 4392c77e088d |
| data/08_source_link_sanitized_recovery/clean_thinking_and_actions__one_per_task.jsonl.gz | Data | 67 B | 1c2d16c635ea |
| isolated_no_guard_manifest.jsonl | Data | 38.4 KB | — |
| manifest.jsonl | Data | 95.5 KB | — |
| official_swebench_pro_verified_baseline_manifest.jsonl | Data | 299.1 KB | — |
| rejected-summary.json | Data | 583 B | — |
| rejected.jsonl | Data | 59.9 KB | — |
| summary.json | Data | 14.3 KB | — |
| swebench_pro_verified_manifest.jsonl | Data | 73.7 KB | — |
| README.md | Documentation | 12.5 KB | — |
| .gitattributes | Repository | 2.5 KB | — |
License and Download
- License
- other
- Access
- No access gate
Released by Zhiyuan Hu through its official repository on Hugging Face.