Clarify SWE LEGO v2 published coverage, completed rerun and missing evidence

swelego_v2/README.md

@@ -2,11 +2,17 @@
 
 Results snapshot: **2026-10-18T21:28:57.295598+00:00**.
 
-## Dataset stats
+## Published dataset stats
 
 - **v1 tasks:** 4,277
 - **Kept:** 1,240 (solved at least once)
-- **Excluded:** 3,037 (no recorded success)
+- **Excluded:** 3,037 (no recorded success in this published snapshot)
+- **Tasks with recorded attempts:** 4,265 / 4,277
+- **Recorded attempts:** 34,120
+- **Available conversations:** 18,582
+- **Missing conversations:** 15,538
+- **Unknown attempt outcomes:** 1,331 across 450 tasks
+- **Tasks with all eight classified outcomes:** 3,815 / 4,277
 
 ## Pass@8 collection
 
@@ -14,9 +20,11 @@ Results snapshot: **2026-10-18T21:28:57.295598+00:00**.
 | --- | ---: | ---: | ---: | ---: |
 | Qwen3-30B Instruct | 4,265 / 4,277 | 34,120 | 3,815 / 4,277 | 1,240 |
 
-The remaining 12 unrun tasks have been resubmitted at pass@8. This snapshot does not yet have complete dataset-wide pass@8 results.
+The table describes the files currently published in this folder. It does not yet provide complete dataset-wide pass@8 results.
 
-### Successful attempts among retained tasks
+**Rerun update (2026-10-20):** The remaining 12 tasks finished on Helma on October 19. All 96 attempts have saved trajectories; five tasks were solved at least once. One attempt ended with `TmuxSessionEndedError`, has no verifier grade, and scores zero under the recorded terminal-error policy. These rerun results are verified locally but are **not yet included in the published task table, traces, or plots**. Incorporating them would bring the kept count to **1,245** and recorded attempts to **34,216**.
+
+### Successful attempts among kept tasks
 
 | Successful attempts | Instruct |
 | --- | ---: |
@@ -33,7 +41,7 @@ The remaining 12 unrun tasks have been resubmitted at pass@8. This snapshot does
 
 `patch.py` rebuilds `tasks.parquet` from v1 at revision `819fe8c3809d9a393502017055c21057b21f02fd` and the accompanying `metadata/selection.json`. The pinned source SHA-256 is verified. Task payloads are current v1 payloads; traces retain the version hashes on which each attempt was evaluated.
 
-31 retained tasks have traces from earlier task versions. The downloadable Parquet contains the current v1 instructions.
+31 kept tasks have traces from earlier task versions. The downloadable Parquet contains the current v1 instructions.
 
 ```python
 import pyarrow.parquet as pq
@@ -42,9 +50,17 @@ subset = pq.read_table("tasks.parquet", filters=[("instruct_successes", "=", 1)]
 
 ## Evidence and trajectories
 
-The combined `traces/` files contain **34,120 unique attempts** from retained and excluded tasks, including the newer recovery results. Saved conversations are available for **18,582** attempts. Original result JSON is preserved where available, alongside the raw verifier reward, outcome reward, and current `retained_v2` flag.
+The combined `traces/` files contain **34,120 unique attempts** from kept and excluded tasks, including the newer recovery results. Saved conversations are available for **18,582** attempts. Original result JSON is preserved where available, alongside the raw verifier reward, outcome reward, and current `retained_v2` flag.
+
+### Missing evidence in the current HF files
+
+- **15,538 attempt rows lack a saved conversation.** Their attempt records are present; a missing conversation does not by itself mean that the verifier outcome is unknown.
+- **1,579 attempts lack a raw verifier grade.** Of these, **248** already have recorded terminal-error evidence and count as reward zero. The other **1,331** have unknown outcomes and no original result JSON in the published trace rows. They remain unknown, rather than being assigned zero without evidence.
+- **12 tasks have no attempt rows in this published snapshot.** Their completed rerun results are described above and await incorporation.
+
+The 1,331 unknown outcomes affect 450 tasks. Together with the 12 rerun tasks not yet incorporated, this leaves 462 tasks without eight classified outcomes in the published snapshot. The historical evidence gaps overlap and should not be added together. Recovery from the original run archives remains pending; the empty fields are not proof that those original traces were lost.
 
-[Trace statistics](trace_stats/README.md) cover the combined results per model, including outcomes, missing evidence, token usage, commands and runtime.
+[Trace statistics](trace_stats/README.md) and plots describe the existing published traces, including their missing evidence. They have not yet been regenerated with the 96 rerun attempts.
 
 ## Runtime