Over-read / Full Read
369,042 tokens
Whole ~100k-character chapter stays in context
SparseRead · arXiv:2608.22237v1
Your agent is burning tokens by reading everything.
SparseRead makes it stop.
Token-efficient reading for AI agents. SparseRead admits only the evidence needed for the next decision, so token volume drops versus Naive full reading — up to 92.9%.
01 / The Problem
Over-reading is object-level access when the current decision needs only a bounded evidence slice. Long-context models, KV-cache tricks, and post-hoc compressors act after a broad object read has already been paid.
In representative high-sparsity tasks, only 6–15% of the exposed reading content contributes to the current decision.
In paired high-sparsity runs, full-read agents spend up to 17.3× more tokens than SparseRead without improving task score.
02 / Frameworks
SparseRead is a thin adapter around one token-efficient reading runtime, not a new agent. The paper evaluates NanoBot, OpenCode, and OpenClaw. Claude Code ships in the GitHub release. Token savings are real and not uniform.
Median 69.0% tokens · 64.4% wall
Median 71.8% tokens · 64.9% wall
Median 28.7% tokens · 28.2% wall
MCP + session hooks · GitHub release
Paper medians cover NanoBot, OpenCode, and OpenClaw · five models · Claude Opus 5 is not in that suite · 12 of 15 cells hold or improve score
03 / SparseRead in Action
Long-context reading / DeepSeek-V4-Flash / NanoBot. Recorded run, labeled REPLAY — not a live model. PaperBench demo tokens are not shown here.
REPLAY of a recorded Figure 8 run — not a live model
Answer five local questions about the Fall of Outremer without reading the whole book into context. LooGLE · DeepSeek-V4-Flash · NanoBot · single run · Figure 8
Over-read / Full Read
369,042 tokens
Whole chapter stays in context
03 write_fileFive answers; score 1.0
SparseRead
50,699 tokens
Five anchored slots, then stop
03 write_fileEvidence State closed; stop. Score 1.0
369,042 → 50,699 tokens · −86.3% · score 1.0 / 1.0 · five local questions, same answers
Single run · SparseRead row is historical same-task in the CSV · Do not present as an average · Figure 8 · results/reports/two_scenario_three_model_baseline_comparison_20260723.csv
04 / How It Works
The control plane decides whether to read sparse. The data plane admits only a bounded slice — backends plug in on one interface. Public terms first; implementation aliases in parentheses.
Object Preview (FileCard) → Read Intent (HintSpec) → Read Gate → macro actions → Evidence State.
The Read Gate switches work modes on its own:force, advisory, or native. Agent-selected macro actions:scout, focus, collect,refine, verify. Evidence State / EvidencePack holds anchors, coverage, unresolved needs, and stop / refine / protocol fallback.
Reader backends do the compression: the object stays outside context until a bounded slice is admitted. Swap Text/PDF, tables, collections, or the script library without changing the Read Gate. A specialized compressor registers as a backend here — not after a full read.
.txt · .md · .pdf · .logfile trees · mixed sources.json · .csv · .tsv · .yaml · .xlsx · .xml.py wrappers · LAMMPS inputFull object
useful evidence
Examples: SWE-Pruner, LLMLingua, or an AI4S domain reader (materials chemistry uses the same snap). SWE-Pruner and LLMLingua are wrapping paths, not evaluated in this work. SparseRead is not post-hoc summarization, KV-cache compression, or an always-on replacement for native tools.
05 / Results
SparseRead is cheaper on tokens and wall time when the task is sparse. Quality usually holds. It is not an always-on win on every cell.
Flagship cell · long-context reading
DeepSeek-V4-Flash on a five-question long document. Score stayed 1.0. One recorded run, not an average.
Strongest model evaluated
On sparse-fit scenarios the strongest model still drops 59.6–89.0% of tokens and 51.3–78.1% of wall time versus full reading.
Same protocol, three harnesses
Median token cut across five models: NanoBot 69.0%, OpenCode 71.8%, OpenClaw 28.7%. Wall-time medians follow the same order.
Against other reading cuts
Versus ACON and last-10 masking, lowest tokens and wall time in every setting. Only method whose score never fell below Naive. 63.4–86.3% tokens; 1.8–3.6× sooner.
Quality usually holds
Four cells dipped between −0.10 and −0.02. Tokens and wall time still fell in all 30. Across the three paper harnesses, 12 of 15 cells hold or improve; three regress between −0.031 and −0.013.
Flagship cell is a single recorded run · Opus range is from the paper, this checkout has no Opus traces · comparison vs ACON and last-10 masking is six settings, some n=1
06 / Quickstart
Source install from GitHub. Clone once, then pick a framework. There is no PyPI or plugin-marketplace pin yet.
git clone https://github.com/Zedong-Liu/SparseReading.git cd SparseReading
Choose a platform
uv pip install \ packages/sparseread-core \ integrations/nanobot/python from sparseread_nanobot import install runtime = install(agent)
Python adapter into the nanobot venv you already have. No installer script.
After install, ask the agent to use SparseRead on a large file. You do not call sro_preview by hand. Full matrix, Windows notes, and doctor checks are in thesource installation guide.
07 / Citation
Read Less, Solve More: Token-Efficient Sparse Reading for AI Agents
arXiv:2608.22237 (v1, 2026-08-23). Setup in the claim registry(125 general tasks from QwenClawBench, Claw-Eval, PinchBench, LooGLE, materials-chemistry AI-for-Science; models Claude Opus 5, Qwen3.6-Plus, DeepSeek-V4-Flash, DeepSeek-V4-Pro, GLM-5.1, Kimi-K2.5); frameworks NanoBot v0.2.0, OpenCode v1.17.14, OpenClaw v2026.6.11. Claude Code 2.1.252 is a fourth GitHub-release integration, not a paper Table 1 row.
This checkout can replay the recorded Figure 8 flagship / multi-file audit / force-mode boundary packets (make reproduce CASE=flagship|strong_model|regime_boundary) and can cite paper ranges. It cannot reproduce from this tree:
make reproduce-suite only checks the three recorded cases)@article{liu2026sparseread,
title={Read Less, Solve More: Token-Efficient Sparse Reading for {AI} Agents},
author={Liu, Zedong and Wu, Jiaan and Ma, Xinyang and Xu, Le and Wang, Kai and Hu, Yuanchao and Tao, Dingwen and Tan, Guangming},
journal={arXiv preprint arXiv:2608.22237},
year={2026},
url={https://arxiv.org/pdf/2608.22237}
}08 / FAQ
A training-free, token-efficient reading layer for AI agents. It admits only the evidence needed for the next decision instead of dumping whole files, PDFs, or repositories into context.
NanoBot, OpenCode, OpenClaw, and Claude Code. One token-efficient reading runtime; four harness adapters.
Clone the GitHub repo, then use the Quickstart for your agent. Source install today — pick a platform and copy the command.
Yes. The unified reader backend can take any new tool — wrap it, register it, and SparseRead will call it instead of reading the whole object first.
Token volume drops by up to 92.9% versus Naive full reading. Wall time drops by up to 89.0%.
09 / Limitations