NNsight interpretability tool runs 70B models remotely on free NSF-backed NDIF GPUs
davidbau · x · 2026-11-19
David Bau highlights NNsight and NDIF as interpretability tooling goes mainstream in production AI. Adding remote = True to a tracing call loads models on NDIF's servers, with hot-swapping for models up to 70B on the free NSF-provided backend; local GPU use is also supported.
- Registration requires US residency or a US-based collaborator; international users without one are pending.
- The no-code NNterp Workbench lets researchers build intervention experiments with clicks and drags.
Related event: NNterp 2.0 Launches with Unified Interpretability API and Tooling Ecosystem(5 posts)→
More from Research
- Epoch AI: Models prefer their own outputs over human gold standard in 23 of 24 tests — BethMayBarnes · 2026-11-19
- tessera-small: a diffusion-LLM-based pixel-art text-to-image model — thejagstudio · 2026-11-19
- WorldRuler: rule-graph world models that honor user-defined game rules — AlayaLab · 2026-11-19
- NexSight: verifiable long-video tasks show top models score only 54.5% — nex-agi · 2026-11-19
- PReMind: benchmark + RL framework for hazard-aware personalized VLM assistants — SeoulNatlUniv · 2026-11-19
- Late Entry in looped LLMs delivers up to 12.5x lossy and 8.4x exact speedups — NanyangTechnologicalUniversity · 2026-11-19