AI & ML interests
Post-training, model behavior and transparent evaluation. Research from alphaBell.
Recent Activity
Articles
ALPHABELL RESEARCH
Experiments you can inspect.
Post-training, model behavior and systems that improve AI.
Five measured model experiments
Change one word, guess the precision, scale a feed-forward branch, follow corrected facts, or hear real routing as music. Explore 60 study generations from Flint and alphaBell Qwen Open, with downloadable evidence and exact CPU reproduction.
Real captures, blind comparisons, activation surgery and model music.
MOSS Layout Studio
Draw text, object and character boxes for MOSS-Image-Base. Keep exact copy in an editable draft, then export the centered model-coordinate JSON.
Runs locally in your browser.
DeepSeek Prompt Parity
Inspect eight saved prompt-format probes for DeepSeek-V4.1-Flash. Compare raw and normalized tool history, numeric reasoning settings, and exact Python/Jinja outputs.
Saved cases with exact inputs and pinned source references.
Flint Repair Clinic
Three real Flint coding failures. Edit the source, inspect the browser error, reveal the recorded repair and play the result.
Real before/after source and hashes. Runs locally in your browser, with no inference call.
Flint Request Doctor
See which settings Rec's Flint MLX server accepts, rejects or silently ignores. Review JSON locally and trace every diagnostic to pinned source.
Flint Arcade
Describe a strange game, play it, then give it one more twist. Stock Flint Q8_0 adapts a tested game foundation, with isolated browser checks and bounded repair.
Each of the three curated games has a recorded winning playthrough using real controls. Explore the code and the generation attempts, including failures.
Application and alphaBell research data under Apache 2.0. Stock model with browser feedback and repair.
Qwen Open 27B
Explore our Qwen3.8-27B research checkpoint, historical coding comparisons and evaluation records.
Noncommercial research. See the model card for lineage, results and release terms.
-
OpenMOSS-Team/MOSS-Image-Base
Text-to-Image • 5B • Updated • 8.56k • 579 -
OpenMOSS-Team/MOSS-Image-SFT
Text-to-Image • 5B • Updated • 13.3k • 774 -
MOSS Layout Studio + Blind Check
🧩Draw MOSS layouts and blind-check two generated outputs
-
MOSS Blind Layout Check
🧪Blind A/B check of MOSS text and placement from local images
-
OpenMOSS-Team/MOSS-Image-Base
Text-to-Image • 5B • Updated • 8.56k • 579 -
OpenMOSS-Team/MOSS-Image-SFT
Text-to-Image • 5B • Updated • 13.3k • 774 -
MOSS Layout Studio + Blind Check
🧩Draw MOSS layouts and blind-check two generated outputs
-
MOSS Blind Layout Check
🧪Blind A/B check of MOSS text and placement from local images
spaces 12
MOSS Layout Studio + Blind Check
Draw MOSS layouts and blind-check two generated outputs
Model Observatory
Flint/Qwen traces and Nanbeige/Qwen architecture
Flint Repair Clinic
Fix three real Flint bugs. Bring the games back to life.
DeepSeek Prompt Parity
Inspect tool history, exact prompts and input contracts
alphaBell Qwen Open Evaluation Explorer
Historical BF16 coding scores and task-level comparisons.