my 0.8B decision model wasn't bad at WinoGrande and ANLI. it was biased.
This is a dev post classified by AGK Intelligence as AI dev tools (a tool drop), kept by the Dev Radar because it carries real work, not commentary.
my 0.8B decision model wasn't bad at WinoGrande and ANLI. it was biased. i looked at what it actually answered, not just the accuracy: it picked the second option 2 out of 3 times, and almost never answered "contradiction", even though a third of the answers were. fixed it with 2,400 class balanced rows. Qwen 3.5 0.8B AA Decision: WinoGrande 50.0 -> 55.0 ANLI 35.5 -> 41.2 BANKING77 56.1 -> 54.2 (the cost) vs Laya, Q4_K_M against their Q8_0: ARC-C - 58.7 vs 31.1 ✅ ARC-E - 77.5 vs 47.9 ✅ BANKING77 - 54.2 vs 33.0 ✅ WinoGrande - 55.0 vs 52.1 ✅ ANLI
Posted by Alok (5.9k followers) 2 days ago · 5 likes · 532 views · view the original post on X. Kept by the Dev Radar as AI dev tools. Tools mentioned: hf.co.
More dev work like this
- 🎃Dario给中国人民带来的极致恐怖从Claude无情封号开始 — @NFT_Chen
- Mia's Qwen3.8-Flash-Next recipe on a DGX Spark: one command, first try, 2 min to the… — @niklaslenz_ai
- 给AI喂一整仓，最烦的就是它每回都从头翻文件，这毛病你多半也遇到过。 — @bkdgiffug
- Coding agents aren't slow because they're dumb. They're slow because every edit waits on… — @RunAnywhereAI
- ICYMI: Foursquare co-founder @dens sat down with our Head of Product, AI, Jeff Smith at… — @automattic
- 🤯惊了！9B参数大模型，只需16G内存即可本地跑通，能力反超Qwen3.5-27B！ — @NFT_Chen
- Introducing JetSpec++: we find causal parallel drafting paired with hidden states… — @haoailab
- Build the agent, test the agent, grade the results, change the agent, run the test cases… — @rseroter
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.2k posts from 5k X accounts over the last 21 days, 8.8k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-10-30 20:27 UTC. Full method.