wassnameandClaudypoo 35be877fc0 pairs: v2 (harder/verbose) + --pairs option; NEGATIVE -- better pairs don't close the 0.67->0.84 gap ...

Authored pairs plateau ~0.67 act / 0.56 grad across all/runtests/allv2; ideal oracle 0.84.
Verbose solutions swamp the localized run_tests hack signal. Pairs lever exhausted.

Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>

2026-06-08 11:53:48 +00:00

..

attic

retract 'null_city contaminated' framing -> in/out-of-subspace + cosine-is-correlational

2026-06-05 09:21:41 +00:00

build_combined_pool.py

reorg: out/ sorted by datatype (vhack/ pools/ runs/ vhack_grads/ figs/)

2026-05-30 03:52:24 +00:00

build_runtests_pool.py

fix: dense run_tests teacher pool (6 -> 215 prompts) so the hack seeds in 60 steps

2026-06-07 11:01:31 +00:00

build_substrate.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

diag_cosine_dist.py

pairs: v2 (harder/verbose) + --pairs option; NEGATIVE -- better pairs don't close the 0.67->0.84 gap

2026-06-08 11:53:48 +00:00

make_random_vhack.py

cleanup: trim 2 stale provenance/train-of-thought comments

2026-06-03 00:25:22 +00:00

pairs_from_rollouts.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

plot_deploy_overlay.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

plot_dynamics.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

plot_emergence.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

plot_substrate.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

probe_distill.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

probe_plot_stack.py

refactor: move 5 leaf entrypoints src/ -> scripts/ (src is now library-only)

2026-06-03 00:23:56 +00:00

rescore_deploy.py

fix: eval on paper test set, not contaminated holdout (base solve 0.94->0.094)

2026-06-07 11:01:31 +00:00

results_deploy.py

results: deploy-eval table (eval2 headline=solve_dep-hack_dep); journal interim read

2026-06-08 10:47:38 +00:00

results.py

fix: results.py parses gt_s/hack_s by header name, not stale fixed indices

2026-05-31 22:45:12 +00:00

tt_erase_bench.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

validate_spoonfeed.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

verify_base_solve.py

fix: eval on paper test set, not contaminated holdout (base solve 0.94->0.094)

2026-06-07 11:01:31 +00:00

verify_eval_gap.py

eval: train/test token gap for all 4 modes (lenient disjoint families)

2026-06-06 13:49:07 +00:00

verify_partition.py

test: no-cheat partition + teacher-pool composition gate (verify_partition.py)

2026-06-05 04:36:03 +00:00

verify_rewards.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

verify_vhack_heldout.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00