CodeSOTA · Experiment · E-PG-1493Portable protocol · public evidence graph
Why this run exists

PR #1493 · SP8192 + 3-Layer Recurrence + Parallel Residuals + Legal TTT

The combined three-layer recurrence, parallel residual routing, QK-gain and legal TTT variant improves the PR #1394 base stack.

Experiment
E-PG-1493
Status
completed
Outcome
supported
Forks
0 public or local starts recorded
Owner
bigbag

Method before metrics.

On PR #1493: 3-layer recurrence, parallel residuals, QK-Gain 5.25, and legal score-first TTT on the PR #1394 stack Declared changes include: SP8192 + 3-layer depth recurrence (L3-5) + parallel residuals (L7+) + QK-Gain 5.25 + legal score-first TTT No SLOT, no pre-quant TTT, no n-gram cache, no ETLB — fully compliant [x] Eval (sliding + TTT) under 600s (~500s actual) [x] Score-first TTT: compliant with Issue #1017 conditions 1-4

Hypothesis
H-PG-1493 · The combined three-layer recurrence, parallel residual routing, QK-gain and legal TTT variant improves the PR #1394 base stack.
Uses technique from
PR #1413 · SP8192 + QK-Gain 5 + Legal Score-First TTT
experiment:E-PG-1413

The comparison contract.

A reproduction matches this protocol. A fork changes it and declares the deviation.

model Family
Parameter Golf language model
dataset
FineWeb validation
benchmarks
FineWeb validation BPB
controlled Variables
16,000,000-byte artifact cap · 600-second training budget · 8×H100 SXM evaluation hardware · tokenizer-agnostic BPB
independent Variable
combined implementation changes declared in the source PR
seeds
42 · 314 · 999
hardware
8×H100 SXM
pr number
1,493
source pr
https://github.com/openai/parameter-golf/pull/1493
source commit
857de4755ef003134e4d3130ca792ff54861c577
source dataset commit
f5c079314c4877fbb0af378c0abade5a8ca33d3a
seed count
3
metric direction
lower_is_better
artifact limit bytes
16,000,000
training budget seconds
600
declared changes
SP8192 + 3-layer depth recurrence (L3-5) + parallel residuals (L7+) + QK-Gain 5.25 + legal score-first TTT · No SLOT, no pre-quant TTT, no n-gram cache, no ETLB — fully compliant · [x] Eval (sliding + TTT) under 600s (~500s actual) · [x] Score-first TTT: compliant with Issue #1017 conditions 1-4 · [x] No SLOT, no pre-quant TTT, no ETLB, no n-gram cache
declared relations
[object Object] · [object Object]

What this experiment produced or evaluated.

Uses
benchmark:fineweb-validation-bpb

Change the evidence, not just the discussion.

Start locally, inspect the portable manifest, and sign in only if you choose to publish.

Attach a run to this protocol.

Record the measurement and its provenance. New evidence is unverified and pending review until CodeSOTA checks it.

Stable, portable, attributable.

Stable ID
E-PG-1493
Visibility
public
Created
2026-04-09
Updated
2026-08-20