Experimental contract
What this family tests
synthetic and PTB repair studies
- Five-seed synthetic architecture test
- Frozen-module PTB block repair
- Local admission-gate ablation
- Joint-gate negative result
Book-to-code crosswalk
The highlighted Experiment cards covered here
- Experiment: PTB-small registered result12.6 A small language-model test · reported result or audit · direct chapter family
Implementation map
From entry point to source surface
Start here
These concrete files are selected from the complete registered source surface to orient the reader.
- runner · public-pinned —
lincs-allora:scripts/run_confirmatory_ket.sh↗Open the pinned public source file. - runner · public-pinned —
lincs-allora:scripts/run_full.sh↗Open the pinned public source file. - runner · public-pinned —
lincs-allora:scripts/run_gpt2_medium_cuda_registered.sh↗Open the pinned public source file. - runner · public-pinned —
lincs-allora:scripts/run_gpt2_medium_cuda_seed7.sh↗Open the pinned public source file.
Author-declared entry points
KET and theory-diagnostic scripts
Resolved code
runner · 37
lincs-allora:KET/__init__.py↗lincs-allora:KET/aggregate_gpt2_cuda_replication.py↗lincs-allora:KET/aggregate_model_zoo.py↗lincs-allora:KET/allora_ket_adapters.py↗lincs-allora:KET/analyze_adapter_sweep.py↗lincs-allora:KET/analyze_capability_safety.py↗lincs-allora:KET/analyze_confirmatory.py↗lincs-allora:KET/analyze_lora_text_sweep.py↗lincs-allora:KET/analyze_theory_diagnostics.py↗lincs-allora:KET/capability_safety_composition.py↗lincs-allora:KET/cross_fitted_quotient_prediction.py↗lincs-allora:KET/factor_quotient_projectability.py↗lincs-allora:KET/gpt2_medium_expanded_eval.py↗lincs-allora:KET/gpt2_medium_mps_pilot.py↗lincs-allora:KET/gpt2_sequential_lora_smoke.py↗lincs-allora:KET/ket_adapter_smoke.py↗lincs-allora:KET/ket_coend_comparison.py↗lincs-allora:KET/ket_ptb_adapter_train.py↗lincs-allora:KET/lora_text_source_smoke.py↗lincs-allora:KET/model_registry.py↗lincs-allora:KET/reflective_allora_order.py↗lincs-allora:KET/stochastic_quotient_aggregation.py↗lincs-allora:scripts/run_confirmatory_ket.sh↗lincs-allora:scripts/run_full.sh↗lincs-allora:scripts/run_gpt2_medium_cuda_registered.sh↗lincs-allora:scripts/run_gpt2_medium_cuda_seed7.sh↗lincs-allora:scripts/run_licket_mps_ket_arch_cross_probe.sh↗lincs-allora:scripts/run_licket_mps_ket_arch_nested_probe.sh↗lincs-allora:scripts/run_licket_mps_learned_split_probe.sh↗lincs-allora:scripts/run_lora_text_mps_big.sh↗lincs-allora:scripts/run_lora_text_mps_lambda_probe.sh↗lincs-allora:scripts/run_lora_text_mps_multiseed.sh↗lincs-allora:scripts/run_lora_text_mps_placement_sweep.sh↗lincs-allora:scripts/run_lora_text_mps_rank_sweep.sh↗lincs-allora:scripts/run_smoke.sh↗lincs-allora:scripts/run_theory_diagnostics.sh↗lincs-allora:scripts/setup_gpt2_composition.sh↗
Evidence record
Artifacts and frozen runs
Result artifacts
repair traces and PTB summaries
Run records
No run-level packet is registered. The family retains code-level provenance, but public run reproduction still requires curation.
Evidence boundary
The failed joint gate and all admission conditions remain visible; accepted local steps do not establish general KET superiority.
This registry links the experiments reported by the three books to their executable or archived code surfaces. A code link establishes traceability, not independent reproduction, correctness, or support for a claim beyond the experiment's stated boundary.