TEMPS: Temporal Sentence Embeddings for Temporal Information Retrieval
Ranking
Overall
82
Content
95
Popularity
N/A
No observed public metrics; popularity remains neutral/archived.
Merged summary
TL;DR - TEMPS adds temporal awareness to frozen semantic retrievers, helping IR and RAG systems find documents that match both the topic and the relevant time. It improves retrieval across three temporal benchmarks without requiring hand-labeled temporal data.
- Introduces Temporal Textual Similarity, which measures time alignment independently of topical similarity.
- Resolves anchored temporal expressions into intervals, represents them as moment-matched Gaussians, and trains an anchor-date-conditioned temporal encoder.
- Fuses temporal and semantic scores at inference while keeping the underlying semantic retriever frozen.
- Improves MRR with every tested semantic backbone and raises TS-Retriever R@1 from 19.92 to 25.39 over the prior temporal state of the art.
Sources (1)
TEMPS: Temporal Sentence Embeddings for Temporal Information Retrieval
Public signals
N/A
TL;DR - TEMPS adds temporal awareness to frozen semantic retrievers, helping IR and RAG systems find documents that match both the topic and the relevant time. It improves retrieval across three temporal benchmarks without requiring hand-labeled temporal data.
- Introduces Temporal Textual Similarity, which measures time alignment independently of topical similarity.
- Resolves anchored temporal expressions into intervals, represents them as moment-matched Gaussians, and trains an anchor-date-conditioned temporal encoder.
- Fuses temporal and semantic scores at inference while keeping the underlying semantic retriever frozen.
- Improves MRR with every tested semantic backbone and raises TS-Retriever R@1 from 19.92 to 25.39 over the prior temporal state of the art.