GT information regimes, KET, and TopoCoend

Experimental contract

What this family tests

controlled comparisons with explicit leakage controls

• Strict-causal versus future-carrier leakage control
• PTB blockwise KET comparison
• WikiText-2 KET diagnostics
• WikiText-103 KET and TopoCoend comparison

Book-to-code crosswalk

The highlighted Experiment cards covered here

1. Experiment: GT information-regime ablation.9.2.3 Revised LM Interpretation · 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 · staged-not-public — categories-gt-archive:KET/ket_experiments/eval/eval_ppl.pyLocate this file in the curated companion packet; no public download is currently offered.
• implementation · public-pinned — lincs-allora:KET/aggregate_gpt2_cuda_replication.py ↗Open the pinned public source file.
• runner · staged-not-public — categories-gt-archive:KET/ket_experiments/eval/freeze_ablate.pyLocate this file in the curated companion packet; no public download is currently offered.
• runner · staged-not-public — categories-gt-archive:KET/ket_experiments/eval/model_utility.pyLocate this file in the curated companion packet; no public download is currently offered.

Author-declared entry points

• KET/ket_experiments/
• scripts/ket_experiments/

Resolved code

Evidence record

Artifacts and frozen runs

Result artifacts

• loss/perplexity plots and recorded comparison tables

Run records

No run-level packet is registered. The family retains code-level provenance, but public run reproduction still requires curation.