Subnet 114 · CoT-Compression-10 · local test bench

Compressor test dashboard

Every batch run on this machine: each compressor arm scored against the baseline (identity) runs of the same batch, with the platform's per-task formula, a 90% bootstrap interval and the stage gates.

Top score
+0.148
s1_fix · project
Current project
+0.148
s1_fix · md5 db938439
Batches
20
none running
Runs recorded
515
402 resolved

Leaderboard

27 arms · click a header to sort
BatchArmCompressor Score90% CIWeighted savingQuality arm / baseTasks RunsUpdated (UTC)Status
gist_s1gistDvariants/gist_novel_D.py-0.031-0.14 … +0.05+0.4%1.00 / 0.885809-28 10:13stage-1 pass
gist_s1gistAvariants/gist_novel_A.py-0.312-0.46 … -0.16-31.7%0.88 / 0.885809-28 10:13stage-1 fail
gist10gistDvariants/gist_novel_D.py––––3009-28 07:48no baseline
gist10gistAvariants/gist_novel_A.py––––3009-28 07:48no baseline
s1_fixprojectbase_miner.py (current project) ★+0.148+0.02 … +0.31+13.8%0.92 / 0.8852509-28 07:46stage-1 pass
s1_gistgistDvariants/gist_novel_D.py––––2009-28 07:45no baseline
s1_gistgistAvariants/gist_novel_A.py––––2009-28 07:45no baseline
gitshow13417projectvariants/base_miner_pre_gitshow_fi––––1309-28 05:28no baseline
gitshow01projectvariants/base_miner_pre_gitshow_fi––––2009-28 05:18no baseline
s1_origprojectvariants/base_miner_pre_gitshow_fi+0.084-0.15 … +0.21+21.2%0.88 / 0.8852509-27 23:17stage-1 pass
ablAbashothervariants/ablation/ablate.py+0.000+0.00 … +0.00-14.8%0.00 / 0.002609-27 04:04stage-1 fail
ablAlistingvariants/ablation/ablate_listing.p+0.000+0.00 … +0.00-27.8%0.00 / 0.002609-27 04:04stage-1 fail
ablBgrephitsvariants/ablation/ablate_grephits.+0.145-0.05 … +0.34+9.1%0.67 / 0.673909-27 03:44stage-1 pass
ablBcommentsvariants/ablation/ablate_comments.-0.135-0.35 … +0.09-22.9%0.67 / 0.673909-27 03:44stage-1 fail
s1_edge10edge10variants/exp_edge10.py-0.094-0.27 … +0.09-18.7%1.00 / 1.005809-26 22:52stage-1 fail
s1_edge10edge10_footervariants/exp_edge10_footer.py-0.033-0.21 … +0.16-19.3%1.00 / 1.005709-26 22:52stage-1 fail
s1_tooluseprojectvariants/base_miner_tooluse_0926.p-0.032-0.12 … +0.09+10.4%1.00 / 1.005809-26 12:56stage-1 pass
s1_gistoffprojectvariants/base_miner_pre_tooluse.py––––56009-26 07:35no baseline
long02notailunsaved version-0.074-0.35 … +0.19+6.4%0.78 / 0.893909-26 03:11stage-1 fail
v4flash01baselineidentity (no compression)––––4009-26 02:48baseline
past01baselineidentity (no compression)––––357009-26 00:46baseline
soma_smokebaselineidentity (no compression)––––1109-25 20:28baseline
capture01g_rereadvariants/g_reread.py––––1109-25 19:41no baseline
long01g_reasoningvariants/base_miner_pre_gitshow_fi-0.119-0.29 … -0.03-23.9%0.87 / 0.9333009-25 16:06stage-1 fail
short01g_reasoningvariants/base_miner_pre_gitshow_fi+0.059-0.05 … +0.16+8.8%1.00 / 1.0052009-25 08:02stage-1 pass
short01g_toolonlyvariants/toolonly.py+0.057-0.04 … +0.15+5.0%1.00 / 1.0052009-25 08:02stage-1 pass
smokeg_reasoningvariants/base_miner_pre_gitshow_fi––––1109-25 06:39no baseline

★ = the current Dev/Compressor/base_miner.py (md5 db938439). Score = platform screener score in [-1, 1]; saving = pooled weighted-token saving per run; quality = resolved share.