🛰️ Daily AI Frontier
‹ back to 2026-09-24

TEMPS: Temporal Sentence Embeddings for Temporal Information Retrieval

Research LLM Agents

Ranking

Overall 82
Content 95
Popularity N/A

No observed public metrics; popularity remains neutral/archived.

Representative image for TEMPS: Temporal Sentence Embeddings for Temporal Information Retrieval

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

arXiv cs.CL Mourad Hassani, Julien Romero, Amel Bouzeghoub, Christian Jacquelinet 2026-09-23 arXiv:2609.28048
Public signals N/A
Providers: Hugging Face · N/A OpenAlex · N/A Publisher · N/A Semantic Scholar · N/A X · N/A Fetched 2026-09-26 14:14:21.519807 UTC

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.
item →