Hi,

It looks like a toolchain configuration inconsistency rather than a TFLM logic issue.

The linker errors suggest that:

Some objects (like tensor_utils.o, instrumentation.o) are built with a different ABI and architecture

While rtthread.elf is using VFP (hardware floating point)

Are you using precompiled TFLM sources, or are they built as part of the RT-Thread project?
Does your project use something like:

-mcpu=cortex-m7
-mfpu=fpv5-sp-d16
-mfloat-abi=hard

If yes, TFLM must match exactly.
geometry dash lite