Discussion on Pre-trained Navigation Policies and Distance Functions
AvivTamar1 · x · 2026-07-10
The conversation revolved around the necessity of additionally training a distance function. A reply pointed out that alongside existing data and the pre-trained CG policy, SPTM merely trains a distance function on the same data. The researchers added that their focus is on improving existing pre-trained GCRL policies, rather than claiming superior overall navigation performance over other methods.
Related event: Research Discusses Improving Pre-trained GCRL Policies(4 posts)→
More from Research
- Open-source runtime lets each repo define its own AI code reviewer — ibabufrik · 2026-07-22
- OpenAI and Apollo Research introduce Contrastive SDF to measure reward-seeking — OpenAI · 2026-07-22
- NVIDIA says to tune the harness before tuning the model with LangChain — NVIDIAAI · 2026-07-22
- The Thimble and the Waterfall: AI's Data Bottleneck and Feedback Loops — dyamins · 2026-07-22
- NVIDIA shows 22 SIGGRAPH papers and Omniverse tools for robot simulation — facontidavide · 2026-07-22
- Building a Knowledge Graph Without a Graph DB: 1000x Cheaper Than GraphRAG — TheRedfather · 2026-07-22