Kimi K3.1 faces hidden-test allegation after reported failure on edited Terminal Bench task

Kimi K3.1 allegedly recalled instructions matching Terminal Bench’s hidden tests, then failed when a task was changed to contradict them. The observer says other models followed the change. A separate commentator doubts the reliability of Kimi’s TB 4.0 score but says the model has been strong in other evaluations. The posts do not establish hidden-test access or provide score documentation.