Skip to content
D-CSIL

Model Registry · HELIX family

HELIX-120M

H-E-R stack · HER×5 (15 layers) · d_model 768 · 12 heads · rotation-HSL (r_max 96) + ElasticAttentionGate + ResonantExpertRouting (MoE top-2 of 4)

STATUS PRIVATEVAL PPL 39.83PROJECT: HELIX
HELIX-120M — TRAINING LOSS (DEMO DATA)152,588 STEPS · MIN 3.68
3.75.57.39.110.90k38k76k114k153k

Benchmarks

MetricValueNotes
Held-out val PPL @ step 152,58839.83corrected number — the in-loop '30.07' was measured val_path==train_path and is invalid
Matched transformer control (same data/steps)33.86 held-outtransformer leads by 5.96 PPL = 15.0%, matched protocol
Open-SLM-style benchmark, best checkpoint38.27%identity-clean-sft-v1 @ step 168,823 — best of any 120M checkpoint
5B-token extensionaborted ×3best held 37.84 @161k then degraded to 39.68 @187k — parameter-bound verdict

Known Limitations

  • The headline foundation val PPL was originally reported as 30.07 — that number is invalid (val_path==train_path, an in-sample measurement); 39.83 is the corrected held-out figure and the one that should ever be cited.
  • A 5B-token continuation was attempted three times; the first two carried an absolute-step warmup bug, and the corrected run still saturated and degraded, indicating the architecture at 127M is parameter-bound, not schedule-bound.
  • Loss-curve anchor points are sparse (only start/end steps evidenced); the true training trajectory is not logged at fine granularity in the registry.
  • Not publicly released — served through a custom, Apple Silicon-only inference engine behind an invite-only gate. There is no public checkpoint, API, or download.