A dataset version is a capture of the source as of a date: its event spine and the files training reads.
2 dataset versions
pooled8 pooled8 @ asof-2026-08-25
the eight streams of the ontology 3.3 pooled control, staged for pooled_v1 on jon-1627: Node's 2026-08-25 capture with the July captures of the other seven, 7,138,794 events by the runs' receipts; members recorded where they are (modal://jon-1627/vei-oss-data/praxis/pooled_v1/streams); see REPORT.md
llm_report · REPORT.md
pooled8@asof-2026-08-25 — data stage report
Recorded 2026-10-13 as part of the OSS line's migration to the factory. The data already existed; nothing was rematerialized or read in bulk.
What it is. The eight streams of the ontology 3.3 pooled control, staged for pooled_v1 on jon-1627: Node's 2026-08-25 capture with the July captures of the other seven, 7,138,794 events by the runs' receipts. A pool is its own source: its runs read it as one body of data, even where a member's bytes are also a single-repo dataset.
Output, by identity. 8 member spines, recorded where they are (modal://jon-1627/vei-oss-data/praxis/pooled_v1/streams/<member>). Each member's digest is the one already on record (its single-repo dataset, or the sync manifest staged beside it); counts and span are copied from the single-repo dataset with the same digest.
Drift. Members drawn from single-repo datasets carry their ingest-time checks there. A pool has no single calendar split (each stream is split at its own time quantile), so no split-time check was run; members without a single-repo dataset of their own have no drift record, a caveat on any result.
the eight streams of the July pooled8 record (drivers_v3), as that run read them: 7,043,088 events, Node's June capture of events created since 2020-01-01 among them; members recorded where they are (archive://Strangelab/ingest-data/oss_events); see REPORT.md
llm_report · REPORT.md
pooled8@asof-2026-07-29 — data stage report
Recorded 2026-10-13 as part of the OSS line's migration to the factory. The data already existed; nothing was rematerialized or read in bulk.
What it is. The eight streams of the July pooled8 record (drivers_v3), as that run read them: 7,043,088 events, Node's June capture of events created since 2020-01-01 among them. A pool is its own source: its runs read it as one body of data, even where a member's bytes are also a single-repo dataset.
Output, by identity. 8 member spines, recorded where they are (archive://Strangelab/ingest-data/oss_events/<member>). Each member's digest is the one already on record (its single-repo dataset, or the sync manifest staged beside it); counts and span are copied from the single-repo dataset with the same digest.
Drift. Members drawn from single-repo datasets carry their ingest-time checks there. A pool has no single calendar split (each stream is split at its own time quantile), so no split-time check was run; members without a single-repo dataset of their own have no drift record, a caveat on any result.
uses the world modelother methodcoin flip / no relationship
AUC is the chance a method ranks a case that happened above one that didn't: 0.5 is a coin flip, 1.0 is perfect.
Spearman ρ is the rank correlation between predicted and measured values: 0 is no relationship, 1.0 is the same order.
The models of record and their headline evaluations. Every result includes the ablations and earlier variants.
apache/airflow: pull requests and contributors (pooled8)
How well the pooled8 world model ranks apache/airflow's pull requests and contributors on what happens next, beside the same facts under a linear head and under gradient-boosted trees.
Contributors likely to return within 90 days
within 90 days
World model + facts (trees)0.749
World model + facts0.749
World model0.718
World model (trees)0.705
Facts (trees)0.745
Facts (linear)0.737
0.450.550.650.750.85
Pull requests likely to get a maintainer's response within 3 days
within 3 days
World model0.621
World model + facts0.621
World model (trees)0.612
World model + facts (trees)0.608
Facts (trees)0.610
Facts (linear)0.610
0.450.550.650.750.85
Pull requests likely to merge
within a week
World model + facts0.839
World model0.832
World model + facts (trees)0.816
World model (trees)0.808
Facts (trees)0.716
Facts (linear)0.706
Clock (trees)0.657
Clock0.459
0.450.550.650.750.85
Pull requests likely to merge
within two weeks
World model + facts0.844
World model0.838
World model + facts (trees)0.826
World model (trees)0.815
Facts (trees)0.720
Facts (linear)0.709
Clock (trees)0.640
Clock0.463
0.450.550.650.750.85
Pull requests likely to merge
within four weeks
World model + facts0.853
World model0.846
World model + facts (trees)0.828
World model (trees)0.820
Facts (trees)0.721
Facts (linear)0.712
Clock (trees)0.626
Clock0.468
0.450.550.650.750.85
Pull requests likely to merge
within three months
World model + facts0.859
World model0.853
World model + facts (trees)0.834
World model (trees)0.822
Facts (trees)0.722
Facts (linear)0.719
Clock (trees)0.614
Clock0.476
0.450.550.650.750.85
How many reviews a pull request still needs
World model + facts0.708
World model0.701
World model + facts (trees)0.693
World model (trees)0.630
Facts (trees)0.650
Facts (linear)0.603
0.000.100.200.300.400.500.600.70
How long a pull request takes to resolve
World model + facts0.647
World model0.644
World model + facts (trees)0.610
World model (trees)0.603
Facts (trees)0.370
Facts (linear)0.292
0.000.100.200.300.400.500.600.70
bitcoin/bitcoin: pull requests and contributors (pooled8)
How well the pooled8 world model ranks bitcoin/bitcoin's pull requests and contributors on what happens next, beside the same facts under a linear head and under gradient-boosted trees.
Contributors likely to return within 90 days
within 90 days
World model + facts (trees)0.781
World model + facts0.769
World model0.749
World model (trees)0.725
Facts (linear)0.766
Facts (trees)0.766
0.450.550.650.750.85
Pull requests likely to get a maintainer's response within 3 days
within 3 days
World model + facts (trees)0.627
World model + facts0.615
World model (trees)0.611
World model0.607
Facts (trees)0.610
Facts (linear)0.596
0.450.550.650.750.85
Pull requests likely to merge
within a week
World model0.857
World model + facts0.857
World model + facts (trees)0.843
World model (trees)0.834
Facts (trees)0.784
Facts (linear)0.783
Clock (trees)0.607
Clock0.485
0.450.550.650.750.85
Pull requests likely to merge
within two weeks
World model + facts0.855
World model0.853
World model + facts (trees)0.850
World model (trees)0.837
Facts (trees)0.785
Facts (linear)0.781
Clock (trees)0.572
Clock0.490
0.450.550.650.750.85
Pull requests likely to merge
within four weeks
World model + facts0.860
World model0.858
World model + facts (trees)0.853
World model (trees)0.841
Facts (trees)0.792
Facts (linear)0.787
Clock (trees)0.545
Clock0.515
0.450.550.650.750.85
Pull requests likely to merge
within three months
World model + facts0.866
World model0.865
World model + facts (trees)0.862
World model (trees)0.855
Facts (trees)0.808
Facts (linear)0.804
Clock0.554
Clock (trees)0.514
0.450.550.650.750.85
How many reviews a pull request still needs
World model + facts (trees)0.759
World model + facts0.728
World model0.726
World model (trees)0.700
Facts (trees)0.671
Facts (linear)0.471
0.000.100.200.300.400.500.600.700.80
How long a pull request takes to resolve
World model + facts0.688
World model0.685
World model + facts (trees)0.646
World model (trees)0.632
Facts (linear)0.551
Facts (trees)0.528
0.000.100.200.300.400.500.600.700.80
tryghost/ghost: pull requests and contributors (pooled8)
How well the pooled8 world model ranks tryghost/ghost's pull requests and contributors on what happens next, beside the same facts under a linear head and under gradient-boosted trees.
Contributors likely to return within 90 days
within 90 days
World model0.814
World model + facts0.760
World model + facts (trees)0.751
World model (trees)0.749
Facts (trees)0.758
Facts (linear)0.734
0.450.550.650.750.85
Pull requests likely to get a maintainer's response within 3 days
within 3 days
World model + facts (trees)0.536
World model0.511
World model (trees)0.488
World model + facts0.472
Facts (trees)0.524
Facts (linear)0.465
0.450.550.650.750.85
Pull requests likely to merge
within a week
World model0.718
World model + facts (trees)0.715
World model + facts0.706
World model (trees)0.695
Facts (trees)0.619
Facts (linear)0.602
Clock0.539
Clock (trees)0.530
0.450.550.650.750.85
Pull requests likely to merge
within two weeks
World model + facts (trees)0.734
World model (trees)0.724
World model0.724
World model + facts0.720
Facts (trees)0.654
Facts (linear)0.615
Clock (trees)0.541
Clock0.540
0.450.550.650.750.85
Pull requests likely to merge
within four weeks
World model + facts (trees)0.749
World model0.741
World model + facts0.739
World model (trees)0.716
Facts (trees)0.666
Facts (linear)0.634
Clock0.540
Clock (trees)0.535
0.450.550.650.750.85
Pull requests likely to merge
within three months
World model + facts0.780
World model + facts (trees)0.767
World model0.766
World model (trees)0.747
Facts (trees)0.691
Facts (linear)0.677
Clock0.559
Clock (trees)0.522
0.450.550.650.750.85
How many reviews a pull request still needs
World model0.650
World model + facts0.642
World model + facts (trees)0.636
World model (trees)0.610
Facts (linear)0.463
Facts (trees)0.461
0.000.100.200.300.400.500.60
How long a pull request takes to resolve
World model0.489
World model + facts0.467
World model + facts (trees)0.437
World model (trees)0.404
Facts (trees)0.313
Facts (linear)0.271
0.000.100.200.300.400.500.60
kubernetes/kubernetes: pull requests and contributors (pooled8)
How well the pooled8 world model ranks kubernetes/kubernetes's pull requests and contributors on what happens next, beside the same facts under a linear head and under gradient-boosted trees.
Contributors likely to return within 90 days
within 90 days
World model + facts0.737
World model + facts (trees)0.732
World model0.689
World model (trees)0.679
Facts (trees)0.738
Facts (linear)0.732
0.350.450.550.650.750.85
Pull requests likely to get a maintainer's response within 3 days
within 3 days
World model + facts (trees)0.633
World model + facts0.622
World model (trees)0.579
World model0.578
Facts (trees)0.637
Facts (linear)0.633
0.350.450.550.650.750.85
Pull requests likely to merge
within a week
World model + facts (trees)0.766
World model + facts0.740
World model (trees)0.714
World model0.701
Facts (trees)0.764
Facts (linear)0.621
Clock (trees)0.602
Clock0.391
0.350.450.550.650.750.85
Pull requests likely to merge
within two weeks
World model + facts (trees)0.789
World model + facts0.750
World model (trees)0.726
World model0.707
Facts (trees)0.771
Facts (linear)0.644
Clock (trees)0.579
Clock0.417
0.350.450.550.650.750.85
Pull requests likely to merge
within four weeks
World model + facts (trees)0.807
World model + facts0.778
World model (trees)0.752
World model0.740
Facts (trees)0.798
Facts (linear)0.680
Clock (trees)0.562
Clock0.451
0.350.450.550.650.750.85
Pull requests likely to merge
within three months
World model + facts (trees)0.880
World model + facts0.860
World model (trees)0.840
World model0.836
Facts (trees)0.862
Facts (linear)0.784
Clock (trees)0.551
Clock0.504
0.350.450.550.650.750.85
How many reviews a pull request still needs
World model + facts0.767
World model + facts (trees)0.761
World model0.760
World model (trees)0.706
Facts (trees)0.685
Facts (linear)0.660
0.000.100.200.300.400.500.600.700.80
How long a pull request takes to resolve
World model + facts0.531
World model0.514
World model + facts (trees)0.507
World model (trees)0.466
Facts (trees)0.456
Facts (linear)0.381
0.000.100.200.300.400.500.600.700.80
NixOS/nixpkgs: pull requests and contributors (pooled8)
How well the pooled8 world model ranks NixOS/nixpkgs's pull requests and contributors on what happens next, beside the same facts under a linear head and under gradient-boosted trees.
Contributors likely to return within 90 days
within 90 days
World model + facts0.629
World model + facts (trees)0.625
World model0.617
World model (trees)0.584
Facts (trees)0.626
Facts (linear)0.618
0.450.550.650.750.85
Pull requests likely to get a maintainer's response within 3 days
within 3 days
World model + facts (trees)0.693
World model (trees)0.669
World model + facts0.638
World model0.635
Facts (trees)0.683
Facts (linear)0.595
0.450.550.650.750.85
Pull requests likely to merge
within a week
World model (trees)0.811
World model + facts (trees)0.810
World model + facts0.805
World model0.800
Facts (trees)0.713
Facts (linear)0.696
Clock (trees)0.672
Clock0.498
0.450.550.650.750.85
Pull requests likely to merge
within two weeks
World model + facts (trees)0.812
World model (trees)0.811
World model + facts0.805
World model0.798
Facts (trees)0.706
Facts (linear)0.692
Clock (trees)0.665
Clock0.513
0.450.550.650.750.85
Pull requests likely to merge
within four weeks
World model + facts (trees)0.808
World model (trees)0.807
World model + facts0.803
World model0.795
Facts (trees)0.697
Facts (linear)0.686
Clock (trees)0.650
Clock0.519
0.450.550.650.750.85
How many reviews a pull request still needs
World model + facts (trees)0.559
World model (trees)0.558
World model + facts0.555
World model0.553
Facts (trees)0.509
Facts (linear)0.469
0.000.100.200.300.400.500.60
How long a pull request takes to resolve
World model + facts0.552
World model0.544
World model (trees)0.532
World model + facts (trees)0.525
Facts (trees)0.313
Facts (linear)0.277
0.000.100.200.300.400.500.60
nodejs/node: pull requests and contributors (pooled8)
How well the pooled8 world model ranks nodejs/node's pull requests and contributors on what happens next, beside the same facts under a linear head and under gradient-boosted trees.
Contributors likely to return within 90 days
within 90 days
World model + facts (trees)0.753
World model + facts0.748
World model0.721
World model (trees)0.707
Facts (linear)0.749
Facts (trees)0.748
0.500.600.700.80
Pull requests likely to get a maintainer's response within 3 days
within 3 days
World model + facts0.756
World model + facts (trees)0.745
World model0.744
World model (trees)0.712
Facts (linear)0.757
Facts (trees)0.754
0.500.600.700.80
Pull requests likely to merge
within a week
World model + facts0.788
World model0.786
World model + facts (trees)0.763
World model (trees)0.758
Facts (trees)0.659
Facts (linear)0.654
Clock (trees)0.580
Clock0.507
0.500.600.700.80
Pull requests likely to merge
within two weeks
World model + facts0.805
World model0.799
World model + facts (trees)0.791
World model (trees)0.781
Facts (trees)0.651
Facts (linear)0.647
Clock (trees)0.548
Clock0.521
0.500.600.700.80
Pull requests likely to merge
within four weeks
World model + facts0.821
World model0.812
World model + facts (trees)0.805
World model (trees)0.793
Facts (trees)0.656
Facts (linear)0.653
Clock0.533
Clock (trees)0.520
0.500.600.700.80
Pull requests likely to merge
within three months
World model + facts0.830
World model0.821
World model + facts (trees)0.810
World model (trees)0.802
Facts (linear)0.651
Facts (trees)0.640
Clock0.550
Clock (trees)0.528
0.500.600.700.80
How many reviews a pull request still needs
World model + facts (trees)0.570
World model + facts0.568
World model0.530
World model (trees)0.501
Facts (trees)0.538
Facts (linear)0.522
0.000.100.200.300.400.500.60
How long a pull request takes to resolve
World model0.456
World model + facts (trees)0.453
World model (trees)0.450
World model + facts0.435
Facts (trees)0.353
Facts (linear)0.317
0.000.100.200.300.400.500.60
opencv/opencv: pull requests and contributors (pooled8)
How well the pooled8 world model ranks opencv/opencv's pull requests and contributors on what happens next, beside the same facts under a linear head and under gradient-boosted trees.
Contributors likely to return within 90 days
within 90 days
World model (trees)0.685
World model0.681
World model + facts (trees)0.670
World model + facts0.643
Facts (trees)0.641
Facts (linear)0.605
0.500.600.700.80
Pull requests likely to get a maintainer's response within 3 days
within 3 days
World model + facts (trees)0.667
World model (trees)0.640
World model + facts0.630
World model0.615
Facts (linear)0.644
Facts (trees)0.633
0.500.600.700.80
Pull requests likely to merge
within a week
World model + facts0.802
World model0.780
World model + facts (trees)0.769
World model (trees)0.750
Facts (linear)0.739
Facts (trees)0.734
Clock (trees)0.544
Clock0.530
0.500.600.700.80
Pull requests likely to merge
within two weeks
World model + facts0.790
World model + facts (trees)0.775
World model0.773
World model (trees)0.750
Facts (linear)0.734
Facts (trees)0.713
Clock (trees)0.531
Clock0.512
0.500.600.700.80
Pull requests likely to merge
within four weeks
World model + facts0.795
World model0.787
World model + facts (trees)0.776
World model (trees)0.753
Facts (linear)0.730
Facts (trees)0.712
Clock (trees)0.521
Clock0.508
0.500.600.700.80
Pull requests likely to merge
within three months
World model0.833
World model + facts0.818
World model + facts (trees)0.792
World model (trees)0.757
Facts (linear)0.750
Facts (trees)0.733
Clock (trees)0.546
Clock0.524
0.500.600.700.80
How many reviews a pull request still needs
World model0.694
World model + facts0.694
World model + facts (trees)0.614
World model (trees)0.582
Facts (linear)0.519
Facts (trees)0.465
0.000.100.200.300.400.500.600.70
How long a pull request takes to resolve
World model + facts0.546
World model0.542
World model + facts (trees)0.494
World model (trees)0.478
Facts (linear)0.362
Facts (trees)0.326
0.000.100.200.300.400.500.600.70
rust-lang/rust: pull requests and contributors (pooled8)
How well the pooled8 world model ranks rust-lang/rust's pull requests and contributors on what happens next, beside the same facts under a linear head and under gradient-boosted trees.
Contributors likely to return within 90 days
within 90 days
World model + facts0.710
World model + facts (trees)0.708
World model0.677
World model (trees)0.647
Facts (linear)0.712
Facts (trees)0.710
0.500.600.700.80
Pull requests likely to get a maintainer's response within 3 days