PathFM learns a joint distribution over future paths conditioned on historical observations and known future covariates.
PathFM generates complete forecast trajectories without sacrificing marginal accuracy
A distilled flow-matching model produces temporally coherent future paths in one or two sampling steps and leads a new trajectory-focused benchmark.
Big Tech
Jiachen Yao · Xiyuan Zhang · Annan Yu · Boran Han · Abdul Fatir Ansari · Shuai Zhang · +2 more
California Institute of Technology · Amazon
Research Digest··2 min read
Yao et al.
Why this paper
From Amazon and California Institute of Technology
In one line
PathFM uses flow matching to generate complete multivariate forecast paths, matching marginal forecast accuracy while preserving trajectory dependence at sampling speeds comparable to quantile models.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§