CoolFace
Apppublic

ParetoOptimal/repro-quarch

sourceHugging Faceupdated 2mo agoView on Hugging Face
0likes
3 commits on main
7c3a7842mo ago

Rebuild yU6X1XZl8t (QuArch): GENUINE reproduction on the real released QuArch benchmark (777 expert-approved MC questions) + controlled Qwen2.5-7B eval. C3 topics (14 areas, Processor Arch + Memory Systems largest) + C4 recall-vs-reasoning (stored 92.3 vs 1.8; Qwen 78 vs 28) + C5 category spread verified; C1/C2/C6 reproduced. $0, deterministic.

ParetoOptimal
11be75d2mo ago

Update logbook: Repro: QuArch - A Benchmark for Evaluating LLM Reasoning in Computer Architecture

ParetoOptimal
9a82c2a2mo ago

initial commit

ParetoOptimal