Unverified50% confidenceFactExact time
Diagnostics on the Sudoku task found that the hidden states of diffusion language models contain erroneous information, but this erroneous information cannot be reliably exposed by token confidence.
1
Sources
50%
Confidence
Long-term
Relevance
11/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOn the Sudoku task, when re-masking an isolated error while keeping the surrounding context unchanged, the regenerated result is often the same error again.76% similarUnverifiedLarge language models still face hallucination problems in information extraction tasks, potentially completing information not explicitly mentioned in conversations and causing inaccurate CRM data.69% similarUnverifiedWhen semantically similar but inaccurate content is retrieved, the model may output incorrect recollections with high confidence, with no warning mechanism at the system level — this is the false memory problem.69% similarUnverifiedFine-tuning a large language model on data that the model itself considers wrong but whose content is harmless can induce broad misalignment.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1166049API
curl https://kongchang.com/api/v1/knowledge/claims/1166049MCP
get_claim(id=1166049)