David Heineman

davidheineman

6 2 1

frimelle's profile picture

andito's profile picture

yjernite's profile picture

https://davidheineman.com

davidheinnman
davidheineman

AI & ML interests

None yet

Recent Activity

updated a model about 21 hours ago

davidheineman/tiny-rl-optimizer-archive

updated a dataset 2 days ago

davidheineman/tiny-rl-refined-tasks-v5.5

published a dataset 2 days ago

davidheineman/tiny-rl-refined-tasks-v5.5

View all activity

Organizations

davidheineman 's collections 14

NanoGPT Ladder v3 - completed pretrained models

57 completed pretrained models, 8M–128M parameters, five data mixtures, 10/30/100 TPP. Original PyTorch weights and generation scripts.

davidheineman/nanogpt-ladder-v3-base-8m-10tpp-seed-1337

Text Generation • Updated 2 days ago • 33
davidheineman/nanogpt-ladder-v3-base-8m-30tpp-seed-1337

Text Generation • Updated 2 days ago • 34
davidheineman/nanogpt-ladder-v3-base-8m-100tpp-seed-1337

Text Generation • Updated 2 days ago • 31
davidheineman/nanogpt-ladder-v3-base-16m-10tpp-seed-1337

Text Generation • Updated 2 days ago • 31

Tiny RL experiment archives

davidheineman/tiny-rl-optimizer-archive

Updated 2 days ago • 929

RLVE Async MOPD N32 Archive (Oct 27 2026)

Cancelled async MOPD student, retained teachers, and verified full-run recovery state.

davidheineman/rlve-async-mopd-n32-cancelled-20261027

Reinforcement Learning • 2B • Updated 17 days ago • 787 • 1

RLVE Multi-round Teachers (Oct 26 2026)

Verified RLVE checkpoints from the Oct 26 storage review, with source provenance and pinned restore revisions.

davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-00-notcontainingstringcounti-773ef32f8f7c

2B • Updated 18 days ago • 33 • 1
davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-04-path-nogoingback-counting-d577b09db95b

2B • Updated 18 days ago • 33 • 1
davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-10-circulatingdecimalcountin-29935810879b

2B • Updated 18 days ago • 33 • 1
davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-01-differenceconstraintsyste-4b8220aaa0a6

2B • Updated 18 days ago • 31 • 1

RLVE Legacy Teachers (Oct 26 2026)

Verified RLVE checkpoints from the Oct 26 storage review, with source provenance and pinned restore revisions.

davidheineman/rlve-archive-mopd-v3-r1-p1r8-teachers-20261004-0950-00-circuit-0ec84b6c0d60

2B • Updated 18 days ago • 39
davidheineman/rlve-archive-mopd-v3-r1-p1r8-teachers-20261004-0950-01-fractionalprogramming-2a02d76e1e2d

2B • Updated 18 days ago • 41 • 1
davidheineman/rlve-archive-mopd-v3-r1-p1r8-teachers-20261004-0950-02-mafmafia-40cec766ee22

2B • Updated 18 days ago • 37
davidheineman/rlve-archive-mopd-v3-r1-p1r8-teachers-20261004-0950-03-cantorexpansion-80be11de127f

2B • Updated 18 days ago • 36

RLVE OPD Teachers (Oct 25 2026)

128 RLVE GRPO teachers from the original data-scaling experiments, initialized from DeepSeek-R1-Distill-Qwen-1.5B.

davidheineman/rlve-opd-teacher-R1Distill-000-NotContainingStringCounting-20261025

Text Generation • 2B • Updated 19 days ago • 116
davidheineman/rlve-opd-teacher-R1Distill-001-DifferenceConstraintSystem-20261025

Text Generation • 2B • Updated 19 days ago • 121
davidheineman/rlve-opd-teacher-R1Distill-002-SumLCM-20261025

Text Generation • 2B • Updated 19 days ago • 122
davidheineman/rlve-opd-teacher-R1Distill-003-MinimumDirectedSpanningTree-20261025

Text Generation • 2B • Updated 19 days ago • 123

RLVE OPD Teachers (R1 Distill Qwen 1.5B)

DeepSeek R1 Distill Qwen 1.5B trained for 150 steps on a single env from https://arxiv.org/abs/2511.07317 (32 of 400 envs)

davidheineman/opd-teacher-R1Distill-FractionalProgramming-step149

2B • Updated Sep 30 • 11
davidheineman/opd-teacher-R1Distill-Axis_KCenter-step149

2B • Updated Sep 30 • 8
davidheineman/opd-teacher-R1Distill-JugPuzzle-step149

2B • Updated Sep 30 • 7
davidheineman/opd-teacher-R1Distill-GaussianElimination-step149

2B • Updated Sep 30 • 10

Verified checkpoint backups - ladder space Nov 2026

Verified checkpoint archives, pinned hashes and restore tools. Local binaries removed after remote verification and CPU restore checks.

davidheineman/ladder-space-20261108-mopd-sweep-n8-rl-20261002-144115-817bc012

Updated 5 days ago
davidheineman/ladder-space-20261108-mopd-sweep-n4-rl-20261002-144115-e02b602c

Updated 5 days ago
davidheineman/ladder-space-20261108-mopd-sweep-n1-rl-20261002-144115-bd5b71e7

Updated 5 days ago
davidheineman/ladder-space-20261108-mopd-v3-r1-n8-targeted-warmstart-20261005-093659-d74fbfe9

Updated 5 days ago

RLVE Async MOPD N16 Replay 30% Archive (Oct 27 2026)

Cancelled async MOPD student, retained teachers, and verified full-run recovery state.

davidheineman/rlve-async-mopd-n16-replay30-cancelled-20261027

Reinforcement Learning • 2B • Updated 17 days ago • 428

RLVE Multi-round Students (Oct 26 2026)

Verified RLVE checkpoints from the Oct 26 storage review, with source provenance and pinned restore revisions.

davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-n16-mopd-fd1ec61ed9c1

2B • Updated 18 days ago • 36
davidheineman/rlve-archive-data-scaling-multiround-r2-2k-v1-20261-n16-mopd-acf365c9c7ec

2B • Updated 18 days ago • 32

RLVE Legacy Students (Oct 26 2026)

Verified RLVE checkpoints from the Oct 26 storage review, with source provenance and pinned restore revisions.

davidheineman/rlve-archive-mopd-v3-r1-all16-warmstart-sft-2026100-sft-all16-9d0226fcb827

2B • Updated 18 days ago • 38 • 1
davidheineman/rlve-archive-hero-longhorizon-d0123-rl-200k-2026101-d0123-rl-200k-8c6fafcfe10f

Updated 18 days ago • 1
davidheineman/rlve-archive-hero-longhorizon-d01-rl-200k-20261015-d01-rl-200k-9cba9fa67528

Updated 18 days ago
davidheineman/rlve-archive-1way-300step-pairwise32-recovery-missi-07-cornfield-91ea3c78398e

2B • Updated 18 days ago • 15

RLVE OPD Students (Oct 25 2026)

8 RLVE MOPD and joint-RL heroes (N16/N32/N64/N128) from the original data-scaling experiments, initialized from DeepSeek-R1-Distill-Qwen-1.5B.

davidheineman/rlve-opd-student-R1Distill-n128-mopd-20261025

Text Generation • 2B • Updated 19 days ago • 126
davidheineman/rlve-opd-student-R1Distill-n128-rl-20261025

Text Generation • 2B • Updated 19 days ago • 121
davidheineman/rlve-opd-student-R1Distill-n16-mopd-20261025

Text Generation • 2B • Updated 19 days ago • 124
davidheineman/rlve-opd-student-R1Distill-n16-rl-20261025

Text Generation • 2B • Updated 19 days ago • 122

RL Data mega-dump (Oct 5 2026)

Dumped all scratch checkpoints from RLVE / nano-omni experiments

davidheineman/rlve-archive-nano-omni-multiteacher-web-85243f9f9dff

Updated Oct 5
davidheineman/rlve-archive-mixing-r1-nofilter16-20260929-231458-00-mix-00-d99a66715866

2B • Updated Oct 5 • 6
davidheineman/rlve-archive-fast-r1-anneal-p4r16-20261003-085736-00-circuit-12de925af98d

Updated Oct 5
davidheineman/rlve-archive-mixing-r1-nofilter16-20260929-231458-01-mix-01-bac4c791a383

2B • Updated Oct 5 • 7

RLVE OPD Teachers (Qwen 2.5 1.5B)

Qwen 2.5 1.5B Instruct trained for 150 steps on a single env from https://arxiv.org/abs/2511.07317 (32 of 400 envs)

davidheineman/opd-teacher-Q2.5I-FractionalProgramming-step149

Text Generation • 2B • Updated Sep 28 • 15
davidheineman/opd-teacher-Q2.5I-Axis_KCenter-step149

Text Generation • 2B • Updated Sep 28 • 14
davidheineman/opd-teacher-Q2.5I-GaussianElimination-step149

Text Generation • 2B • Updated Sep 28 • 17
davidheineman/opd-teacher-Q2.5I-JugPuzzle-step149

Text Generation • 2B • Updated Sep 28 • 17

NanoGPT Ladder v3 - completed pretrained models

57 completed pretrained models, 8M–128M parameters, five data mixtures, 10/30/100 TPP. Original PyTorch weights and generation scripts.

davidheineman/nanogpt-ladder-v3-base-8m-10tpp-seed-1337

Text Generation • Updated 2 days ago • 33
davidheineman/nanogpt-ladder-v3-base-8m-30tpp-seed-1337

Text Generation • Updated 2 days ago • 34
davidheineman/nanogpt-ladder-v3-base-8m-100tpp-seed-1337

Text Generation • Updated 2 days ago • 31
davidheineman/nanogpt-ladder-v3-base-16m-10tpp-seed-1337

Text Generation • Updated 2 days ago • 31

Verified checkpoint backups - ladder space Nov 2026

Verified checkpoint archives, pinned hashes and restore tools. Local binaries removed after remote verification and CPU restore checks.

davidheineman/ladder-space-20261108-mopd-sweep-n8-rl-20261002-144115-817bc012

Updated 5 days ago
davidheineman/ladder-space-20261108-mopd-sweep-n4-rl-20261002-144115-e02b602c

Updated 5 days ago
davidheineman/ladder-space-20261108-mopd-sweep-n1-rl-20261002-144115-bd5b71e7

Updated 5 days ago
davidheineman/ladder-space-20261108-mopd-v3-r1-n8-targeted-warmstart-20261005-093659-d74fbfe9

Updated 5 days ago

Tiny RL experiment archives

davidheineman/tiny-rl-optimizer-archive

Updated 2 days ago • 929

RLVE Async MOPD N16 Replay 30% Archive (Oct 27 2026)

Cancelled async MOPD student, retained teachers, and verified full-run recovery state.

davidheineman/rlve-async-mopd-n16-replay30-cancelled-20261027

Reinforcement Learning • 2B • Updated 17 days ago • 428

RLVE Async MOPD N32 Archive (Oct 27 2026)

Cancelled async MOPD student, retained teachers, and verified full-run recovery state.

davidheineman/rlve-async-mopd-n32-cancelled-20261027

Reinforcement Learning • 2B • Updated 17 days ago • 787 • 1

RLVE Multi-round Students (Oct 26 2026)

Verified RLVE checkpoints from the Oct 26 storage review, with source provenance and pinned restore revisions.

davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-n16-mopd-fd1ec61ed9c1

2B • Updated 18 days ago • 36
davidheineman/rlve-archive-data-scaling-multiround-r2-2k-v1-20261-n16-mopd-acf365c9c7ec

2B • Updated 18 days ago • 32

RLVE Multi-round Teachers (Oct 26 2026)

Verified RLVE checkpoints from the Oct 26 storage review, with source provenance and pinned restore revisions.

davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-00-notcontainingstringcounti-773ef32f8f7c

2B • Updated 18 days ago • 33 • 1
davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-04-path-nogoingback-counting-d577b09db95b

2B • Updated 18 days ago • 33 • 1
davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-10-circulatingdecimalcountin-29935810879b

2B • Updated 18 days ago • 33 • 1
davidheineman/rlve-archive-data-scaling-multiround-r2-v1-20261023-01-differenceconstraintsyste-4b8220aaa0a6

2B • Updated 18 days ago • 31 • 1

RLVE Legacy Students (Oct 26 2026)

Verified RLVE checkpoints from the Oct 26 storage review, with source provenance and pinned restore revisions.

davidheineman/rlve-archive-mopd-v3-r1-all16-warmstart-sft-2026100-sft-all16-9d0226fcb827

2B • Updated 18 days ago • 38 • 1
davidheineman/rlve-archive-hero-longhorizon-d0123-rl-200k-2026101-d0123-rl-200k-8c6fafcfe10f

Updated 18 days ago • 1
davidheineman/rlve-archive-hero-longhorizon-d01-rl-200k-20261015-d01-rl-200k-9cba9fa67528

Updated 18 days ago
davidheineman/rlve-archive-1way-300step-pairwise32-recovery-missi-07-cornfield-91ea3c78398e

2B • Updated 18 days ago • 15

RLVE Legacy Teachers (Oct 26 2026)

Verified RLVE checkpoints from the Oct 26 storage review, with source provenance and pinned restore revisions.

davidheineman/rlve-archive-mopd-v3-r1-p1r8-teachers-20261004-0950-00-circuit-0ec84b6c0d60

2B • Updated 18 days ago • 39
davidheineman/rlve-archive-mopd-v3-r1-p1r8-teachers-20261004-0950-01-fractionalprogramming-2a02d76e1e2d

2B • Updated 18 days ago • 41 • 1
davidheineman/rlve-archive-mopd-v3-r1-p1r8-teachers-20261004-0950-02-mafmafia-40cec766ee22

2B • Updated 18 days ago • 37
davidheineman/rlve-archive-mopd-v3-r1-p1r8-teachers-20261004-0950-03-cantorexpansion-80be11de127f

2B • Updated 18 days ago • 36

RLVE OPD Students (Oct 25 2026)

8 RLVE MOPD and joint-RL heroes (N16/N32/N64/N128) from the original data-scaling experiments, initialized from DeepSeek-R1-Distill-Qwen-1.5B.

davidheineman/rlve-opd-student-R1Distill-n128-mopd-20261025

Text Generation • 2B • Updated 19 days ago • 126
davidheineman/rlve-opd-student-R1Distill-n128-rl-20261025

Text Generation • 2B • Updated 19 days ago • 121
davidheineman/rlve-opd-student-R1Distill-n16-mopd-20261025

Text Generation • 2B • Updated 19 days ago • 124
davidheineman/rlve-opd-student-R1Distill-n16-rl-20261025

Text Generation • 2B • Updated 19 days ago • 122

RLVE OPD Teachers (Oct 25 2026)

128 RLVE GRPO teachers from the original data-scaling experiments, initialized from DeepSeek-R1-Distill-Qwen-1.5B.

davidheineman/rlve-opd-teacher-R1Distill-000-NotContainingStringCounting-20261025

Text Generation • 2B • Updated 19 days ago • 116
davidheineman/rlve-opd-teacher-R1Distill-001-DifferenceConstraintSystem-20261025

Text Generation • 2B • Updated 19 days ago • 121
davidheineman/rlve-opd-teacher-R1Distill-002-SumLCM-20261025

Text Generation • 2B • Updated 19 days ago • 122
davidheineman/rlve-opd-teacher-R1Distill-003-MinimumDirectedSpanningTree-20261025

Text Generation • 2B • Updated 19 days ago • 123

RL Data mega-dump (Oct 5 2026)

Dumped all scratch checkpoints from RLVE / nano-omni experiments

davidheineman/rlve-archive-nano-omni-multiteacher-web-85243f9f9dff

Updated Oct 5
davidheineman/rlve-archive-mixing-r1-nofilter16-20260929-231458-00-mix-00-d99a66715866

2B • Updated Oct 5 • 6
davidheineman/rlve-archive-fast-r1-anneal-p4r16-20261003-085736-00-circuit-12de925af98d

Updated Oct 5
davidheineman/rlve-archive-mixing-r1-nofilter16-20260929-231458-01-mix-01-bac4c791a383

2B • Updated Oct 5 • 7

RLVE OPD Teachers (R1 Distill Qwen 1.5B)

DeepSeek R1 Distill Qwen 1.5B trained for 150 steps on a single env from https://arxiv.org/abs/2511.07317 (32 of 400 envs)

davidheineman/opd-teacher-R1Distill-FractionalProgramming-step149

2B • Updated Sep 30 • 11
davidheineman/opd-teacher-R1Distill-Axis_KCenter-step149

2B • Updated Sep 30 • 8
davidheineman/opd-teacher-R1Distill-JugPuzzle-step149

2B • Updated Sep 30 • 7
davidheineman/opd-teacher-R1Distill-GaussianElimination-step149

2B • Updated Sep 30 • 10

RLVE OPD Teachers (Qwen 2.5 1.5B)

Qwen 2.5 1.5B Instruct trained for 150 steps on a single env from https://arxiv.org/abs/2511.07317 (32 of 400 envs)

davidheineman/opd-teacher-Q2.5I-FractionalProgramming-step149

Text Generation • 2B • Updated Sep 28 • 15
davidheineman/opd-teacher-Q2.5I-Axis_KCenter-step149

Text Generation • 2B • Updated Sep 28 • 14
davidheineman/opd-teacher-Q2.5I-GaussianElimination-step149

Text Generation • 2B • Updated Sep 28 • 17
davidheineman/opd-teacher-Q2.5I-JugPuzzle-step149

Text Generation • 2B • Updated Sep 28 • 17