Topic

Inference determinism

Why identical prompts to the same model can yield different outputs, covering decoding settings, seeds, floating-point non-associativity on GPUs and request batching on hosted endpoints.

Current clusters

build2 publishers

Five retrieved guidelines lift AppWorld's hard-task success by 14.2 points

IBM Research's ALTK-Evolve distils an agent's own trajectories into scored guidelines and injects the top five at inference time, and a companion post puts a number on the reliability an average success rate hides.

Publishers:dev.tohuggingface.co

Reality

Evidence45
Adoption20
Hype gap+15
Incentives85
Confidence55