Unverified50% confidenceFactExact time
A study on Bengali idioms used six frozen encoders, 10,359 Bengali idiom items, and eight input conditions to systematically test the reading, transfer, and erasure of cross-lingual concept subspaces.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA study on Bengali coreference resolution systematically evaluated multiple settings, including zero-shot, fine-tuning, data augmentation and cross-lingual transfer.71% similarUnverifiedMany FLORES sentences are not in Unicode NFC normalized form: 609 of 1,012 Bengali sentences, 573 Punjabi sentences, and 93 Hindi sentences; failing to normalize first causes token count drift.70% similarUnverifiedThe study proposes routing linguistic and paralinguistic information into different representation channels to achieve controllable disentanglement without retraining the encoder.68% similarUnverifiedA study based on a corpus containing 131,000 sentences covering 11 Indian scripts compared the tokenizer performance of models such as TinyLlama-1.1B, mBERT, and MuRIL.67% similarUnverifiedA new study for the first time extends an experimental framework for evaluation awareness to Indian languages, testing four instruction-tuned models from three model families, covering five Indian languages and English, and producing and publicly releasing a total of 47,880 judged responses.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1153325API
curl https://kongchang.com/api/v1/knowledge/claims/1153325MCP
get_claim(id=1153325)