Back to corpus
proposalexperiment writeup candidatescore 30

N'Ko Research Papers

The public closeout series now lives under `final/`. Each paper has its own folder with a local `paper.tex`, compiled `paper.pdf`, `references.bib`, `paper.bbl`, and relative `figures/` assets, so each manuscript can compile from its own directory.

Full HTML reader

Read the full artifact

Open in new tab

Extracted abstract or opening context

The public closeout series now lives under `final/`. Each paper has its own folder with a local `paper.tex`, compiled `paper.pdf`, `references.bib`, `paper.bbl`, and relative `figures/` assets, so each manuscript can compile from its own directory. | # | Paper | Folder | PDF pages | Role | |---|-------|--------|-----------|------| | 1 | Dead Circuits: Script Invisibility and Representation Failure for N'Ko in Large Language Models | `final/01-script-invisibility/` | 11 | Establishes the LLM representation failure with research questions, falsification criteria, evidence ladder, tokenizer-burden formalism, evidence artifact contract, remediation agenda, reviewer checklist, and validity threats. | | 2 | Against WER: Phonemic Evaluation, Orthographic Transparency, and the Script Advantage for Manding ASR | `final/02-phonemic-evaluation/` | 10 | Formalizes the metric problem, N'Ko-vs-Latin script advantage, transparent-script edit preservation, normalization protocol, CER/PER proxy boundary, metric failure taxonomy, and matched-evaluation requirements. | | 3 | Script-Native ASR for N'Ko: Anticipatory Transformer CTC Decoding and the 20.57% CER Anchor | `final/03-script-native-asr-anchor/` | 12 | Preserves the technical ASR anchor with architecture, trajectory-state math, corpus split, 20.57% CER arithmetic, hashes, allowed/disallowed claims, provenance notes, artifact contract, operational lessons, and model-card boundaries. | | 4 | Anticipation Geometry Partition: Row-Level Governance for Script-Native N'Ko ASR Deployment | `final/04-agp-deployment/` | 12 | Defines AGP as the post-ASR correction/provenance/deployment governance layer with system boundaries, pipeline formalization, row contracts, partition scoring, correction benchmark design, failure taxonomy, human review, data lifecycle, Djoko substrate, and ExpF/ExpH evidence. | Recommended public narrative: use the four papers as the final publishable bundle and treat `current/paper_canonical_nko_agp_20cer.tex` as the synthesis manuscript that explains how the four papers connect. The 20.57% CER should be described as an archived N'Ko trajectory ASR checkpoint under recorded settings, not as a universal matched proof against Latin. The readable public companion lives in `blog-series/`. It inherits the stronger voice from the original `blog/posts/` drafts: historical opening, experiment chronology, concrete numbers, and plain-English explanations before acronyms. Start with `blog-series/00-field-guide-to-the-claim.md`, then publish the four essays in order. | # | Paper | File | Status | Pages | |---|-------|------|--------|-------| | 1 | Dead Circuits: Activation Profiling and Script Invisibility in LLMs | `current/paper1_dead_circuits.tex` | Draft complete | ~20 | | 2 | Living Speech: Script-Nat

Promotion decision

What has to happen next

Attach run IDs, datasets, metrics, and reproduction commands.

Why this is not always a full paper yet

Corpus pages are public-safe readers for discovered workspace artifacts. They are not automatically final papers. A corpus item becomes a polished paper only after the editable source, evidence checkpoints, references, figures, render path, and release status are attached through the paper schema.