Read Less, Solve More: Token-Efficient Sparse Reading of External Content for AI Agents
Paper · https://arxiv.org/abs/2608.22237
Web-facing AI agents repeatedly access webpages, documents, repositories, and tool responses, yet reading interfaces often expose entire objects when only sparse evidence is needed. This over-reading increases token and latency costs and can dilute relevant evidence. We present SparseRead, a reusable reading control layer that governs content admission without modifying the agent model. Its Read Intent Protocol tracks evolving needs and source coverage; Adaptive Reader Backends select how evidence is located and exposed; and the Budgeted Read Gate bounds cumulative raw exposure per episode with faithful full-view fallback. A unified Reader interface supports interchangeable evidence acquisition across heterogeneous resources. Across six models, including Claude Opus 5, and five scenarios, SparseRead reduces total tokens by up to 92.9% and wall time by up to 89.0%, preserving or improving task quality in 97% settings. On Web deep-search tasks, it reduces tokens by 31.0--43.0% and task completion time by 23.0--37.0%. Ablations validate the three designs, while integration with three agent frameworks demonstrates portability.
First seen by Oriel on .
AI summary
SparseRead is a model-independent reading control layer that selectively admits external content, reducing agent token use and latency while preserving or improving task quality.
- Its Read Intent Protocol tracks evolving information needs and source coverage during an episode.
- Adaptive Reader Backends locate and expose evidence across webpages, documents, repositories, and tool responses through a unified Reader interface.
- A Budgeted Read Gate bounds cumulative raw exposure while retaining a faithful full-view fallback.
- Across six models and five scenarios, SparseRead cuts tokens by up to 92.9% and wall time by up to 89.0%.
- Task quality was preserved or improved in 97% of settings, while Web deep-search tokens fell 31.0–43.0% and completion time 23.0–37.0%.
#ai agents #information retrieval #token efficiency #web search
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Points
Score reported by the source.
- Hugging Face Papers: 24, latest observation ; +2 over observed period since
Comments
Comment count reported by the source.
- Hugging Face Papers: 12, latest observation ; 0 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Papers | 2608.22237 | https://arxiv.org/abs/2608.22237 Read discussion |