arbiter:4brejects 57 %
Arbiter · Codekins · Zyot Lab · 4.3B · ONNX Runtime · Apache-2.0 (LoRA adapter and head) and the Gemma Terms of Use (Gemma 3 base model)
- Score
- 15.2
- 95 % interval 13.2–17.1
- Rank
- 31
- of 35 scored models
- Public benchmark
- 42.0
- macro accuracy, %
- typed-decisions
- 0.0
- accuracy, %
By task
Chance-corrected accuracy on each of the 14 tasks: 0 is guessing, 100 is every answer right.
Accuracy and calibration
| Benchmark | Accuracy | Choice | Noul | Score | ECE | Brier | Rejected |
|---|---|---|---|---|---|---|---|
| Public benchmark | 42.0(40.9–43.2) | 38.1 | 71.0 | 0.0 | 0.137 | 0.475 | 1,333 / 3,880 |
| typed-decisions | 0.0(0.0–0.0) | 0.0 | 0.0 | 0.0 | – | – | 400 / 400 requests |
Accuracy in %, with its 95 % bootstrap interval. A rejected request (the model can’t take the question, or the input is too long for it) counts as wrong.
Speed and memory
| Device | p50 | p90 | Load | VRAM | RAM peak | Precision |
|---|---|---|---|---|---|---|
| RTX 5090 CUDA | 135 ms | 135 ms | 6.79 s | 8.3 GB | 2.0 GB | F32 |
| Threadripper 3960X CPU | 6.11 s | 6.13 s | 6.57 s | – | 7.2 GB | F32 |
Five-question requests over HTTP, one at a time, each a different short message. VRAM: the rise of the GPU’s used memory while the model is loaded. RAM peak: the runner process’s peak resident memory.
Parity with the authors’ code
Its parity check is in the model’s family documentation in the repository (docs/families/); the compiled parity file does not list it yet.
Reproduce
The tools are in convert/ollaya_convert and need only Python’s standard library. Run them against a local Ollaya 0.13.0 with the same model digest.
ollaya pull arbiter:4b
# typed-decisions; export the test split once with
# uv run python -m ollaya_convert.bench_typed export --out typed.jsonl
python -m ollaya_convert.bench_typed run --model arbiter:4b \
--url http://127.0.0.1:11435 --data typed.jsonl --out typed/arbiter-4b
# Bespoke Labs' public benchmark; build its 13 subsets first
NIMBLE_SRC=nimble python -m ollaya_convert.bench_public run --model arbiter:4b \
--url http://127.0.0.1:11435 --data public --out public/arbiter-4b
# speed and memory
python -m ollaya_convert.bench_latency --models arbiter:4b --memory \
--machine my-machine --device cuda --out latency.json