I've been thinking about the impact of AI on education and spent some time trying out the latest AI detection tools this weekend.
linkedin.com/pulse/pitfallsโฆ
linkedin.com/pulse/pitfallsโฆ
@samrmadden
Recent agentic systems (Claude Code, Codex, RLM, etc.) push context out of the prompt and into the environment (e.g., as files). This helps them maintain long-term knowledge about their goals and functionality.View quoted post on X
๐จ While this is a good idea, we show a surprising result: systems that use external environments like this perform much better when given a small, fixed-size, in-context, agent-managed cache that "๐ฑ๐ฆ๐ฆ๐ฌ๐ด ๐ช๐ฏ๐ต๐ฐ" these environments.
๐ Our paper, ๐ฃ๐๐๐: ๐ ๐จ๐ฎ๐จ๐ฉ๐๐ข ๐๐ค๐ง ๐๐ช๐๐ก๐๐๐ฃ๐ ๐๐ฃ๐ ๐ข๐๐๐ฃ๐ฉ๐๐๐ฃ๐๐ฃ๐ ๐ฎ๐ป ๐ผ๐ฟ๐ถ๐ฒ๐ป๐๐ฎ๐๐ถ๐ผ๐ป ๐ฐ๐ฎ๐ฐ๐ต๐ฒ ๐๐ค๐ง ๐๐๐ ๐๐๐๐ฃ๐ฉ๐จ, introduces this idea.
Compared with strong baselines, including RAG, Compaction Agents, and SOTA prompt-learning frameworks, PEEK dominates the costโquality Pareto frontier: achieving +6.3โ34.0% in quality, with fewer iterations and lower cost.
Paper: arxiv.org/abs/2605.19932
GitHub: github.com/zhuohangu/peek
More in the thread below! (1/N)
๐ฃ We're spreading the word about #SemBench -- a brand new benchmark for semantic query processing over multimodal workloads including text, image, audio, and tabular data!View quoted post on X
๐Paper: bit.ly/3WGsZf6
๐ปWebsite: sembench.org
๐พCode: bit.ly/49DeuAg
๐ข CfP: Conference on Innovative Data Systems Research #CIDR2026 (Chaminade, USA). Papers due: August 5, 2025. Gustavo Alonso, @samrmadden, and I are looking forward to your paper & demo submissions! cidrdb.org/cidr2026/View quoted post on X
The CIDR 2025 program is online: cidrdb.org/cidr2025/progrโฆView quoted post on X
Registration closes today -- Join us in Amsterdam:
cidrdb.org/cidr2025/regisโฆ