Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.
Python calibrationclassificationdecision-modelhuggingfacejevmodernbertmultilingualnlppythonpytorchroutingtyped-decisionszero-shot
chore: keep the benchmark's frozen bytes LF
GaotianJin46257d9c18be
diff --git a/.gitattributes b/.gitattributesindex f491c0bf..f09fdcb2 100644--- a/.gitattributes+++ b/.gitattributes@@ -6,3 +6,10 @@research/benchmarks/feishu_zh/SOURCE.json text eol=lfresearch/benchmarks/feishu_zh/data/** text eol=lfresearch/benchmarks/feishu_zh/results/** text eol=lf++# Same reason in research/benchmarks/zh_short_commands: audit.py pins the sha256 of the+# raw bytes of data/cases.jsonl and prompts.py, and a CRLF report.json would make every+# later run on another platform look like a rewrite.+research/benchmarks/zh_short_commands/data/** text eol=lf+research/benchmarks/zh_short_commands/prompts.py text eol=lf+research/benchmarks/zh_short_commands/results/** text eol=lf