Skip to content

docs(TODO.impl/11): framing verdict — node-kernel finding #97

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ronaldtse merged 1 commit into from
Sep 12, 2026
Merged

docs(TODO.impl/11): framing verdict — node-kernel finding #97

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 23 additions & 2 deletions TODO.impl/11-static-int8-framing.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,10 +37,31 @@ If A and B clear: re-export all quantized artifacts through the static
path (modal_export gains the calibration stage), index-v6, golden rows
regenerate, framing claim added to the contract.

## Verdict (2026-09-12, final): framing gate FAILED for the node fix — the instability is the node build itself

The test run at every endpoint (RESULTS.md, PR #214):

- Python ORT: 480/480 framing-stable at fp32, dynamic int8, AND
static int8
- onnxruntime-node: 89-101/485 UNSTABLE at every precision INCLUDING
fp32 — batch-shape-dependent kernel numerics before quantization
enters; static scales cannot repair what is not a scale computation
- static int8 byproduct: quality-clean on sample (0/480 drift vs
fp32) and ~8% faster than dynamic on CPU — stays a candidate for
the export path on its own merits, gated by full-set DER, NOT by
framing

Re-scoped conclusion: framing stability is a property of the runtime
distribution. Shipped TS paths are single-framing by construction, so
published byte parity is unaffected; framing is the parity variable
of mixed-shape serving (why the speculative tier degraded and was
pulled — TODO.impl/01).

## Status

- [x] Pre-registered (this file)
- [x] fp32 artifact fetched, sha-verified against index-v5
- [x] Calibration corpus collected (365 feeds, both framings)
- [ ] quantize_static result measured (A/B/C)
- [ ] Verdict recorded in RESULTS.md; ledger cross-links updated
- [x] quantize_static measured (A/B/C) across BOTH runtime builds
- [x] Verdict recorded in RESULTS.md + Paper C 5.1; framing-fix
closed negative; static-int8 speed/quality kept open
2 changes: 1 addition & 1 deletion TODO.impl/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ Priorities:
| 08 | golden-closeout.md | golden test PR, paper-c wording, release README note | P0 | closed 2026-09-12 (PR #211 merged) |
| 09 | doc-corrections.md | Engram mechanism fix, mHC single-pass note | P2 | landed in PR #95 |
| 10 | engram-lexical-memory.md | hashed n-gram memory module | P3 | gated behind 04 |
| 11 | static-int8-framing.md | static activation scales: fix framing at the source | P0 | pre-registered; experiment running |
| 11 | static-int8-framing.md | static scales as the framing fix | P0 | closed: node-kernel finding; static stays open on speed |

Standing rules that apply to every item: full-set measurement or no
claim; quantized parity is quality-level, not byte-level; LLM teachers
Expand Down