Unverified50% confidenceFactExact time
Marimba's turn-end judgment is read from layer-18 features of the backbone network, and the speaker-related readout is extracted from layer-12 features.
1
Sources
50%
Confidence
Long-term
Relevance
11/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTokka-Bench proposes five evaluation dimensions: bytes per token, unique token coverage, fertility, word split rate, and vocabulary composition.58% similarUnverifiedThe cognitive levels in Bloom's Taxonomy include dimensions such as remembering, understanding, applying and analyzing, which can be used for metadata annotation of questions in a question bank.56% similarUnverifiedLexRank constructs an undirected graph using TF-IDF cosine similarity and measures sentence importance with spectral graph centrality.56% similarUnverifiedThe tool's "Grammar Analysis" feature uses levels such as L0 and L1 to annotate the nesting relationships between main clauses and subordinate clauses, and identifies the subject, predicate, and object at each level.56% similarUnverifiedThe textual description of the leaked feature was semantically correct, but the code implementation read future candlestick data, which is why it passed review.56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1173841API
curl https://kongchang.com/api/v1/knowledge/claims/1173841MCP
get_claim(id=1173841)