Unverified50% confidenceFactExact time
In Marimba's readout logic, vocabulary-type questions are scored zero-shot directly on the ASR-transcribed text, with no additional classifier training required.
1
Sources
50%
Confidence
Long-term
Relevance
11/16/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedZero-shot classification relies on the semantic understanding capabilities of pretrained large language models such as NLI variants of BART and RoBERTa.63% similarUnverifiedThe working mechanism of Text Classifier is essentially zero-shot/few-shot classification—the model does not need special training and only needs to read the category descriptions to make a judgment, so the clarity of the category descriptions directly determines classification accuracy.62% similarUnverifiedByte-level language models use raw byte sequences directly as the input and output units, without relying on a pretrained tokenizer.60% similarUnverifiedVulnerability fixed in 3.0.1: claims comparing ASTs or bytecode could still pass via the comment canary; subsequently the canary added no-op assignments to change the syntax tree and bytecode, but this was a partial fix, and claims that read only a smaller fragment of the syntax tree could still pass.60% similarUnverifiedThe fact that full transcripts yield lower accuracy than no notes indicates the problem lies in the discussion-generation stage itself rather than in memory compression.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1173839API
curl https://kongchang.com/api/v1/knowledge/claims/1173839MCP
get_claim(id=1173839)