Paper v110 Nov
Time window
Sort by
How Attention is calculated
What the score means. Attention is a relative 0–100 estimate of how much public interest a benchmark is receiving—or, for a brand-new release, may receive. It does not measure benchmark quality or model performance.
- Latest: 45% Hugging Face paper votes + 25% GitHub stars + 5% dataset downloads, with up to 25 bonus points from the experimental LLM 7-day forecast.
- 30 days: 20% paper votes + 62.86% GitHub stars + 17.14% dataset downloads. No LLM forecast.
- 90 days: 5% paper votes + 61.47% GitHub stars + 33.53% dataset downloads. No LLM forecast.
How the forecast works. The LLM considers only the supplied benchmark scope, artifact readiness, topical breadth, and verified publisher evidence. It does not invent stars, votes, downloads, or adoption.
Normalization and missing data. Latest uses within-batch percentiles; 30- and 90-day counts use log normalization against the window maximum. A missing signal receives a neutral value of 0.5—not zero—and lowers confidence. Scores use the last known observations; the page update date does not mean every signal was freshly collected.
Rising benchmarks
14 resultsYour interest list
Choose the directions you want to keep up with. Your list stays here for your next visit.
Could not save preferences. Your selection applies to this visit.
Paper v110 Nov
Chem World
Molecular property prediction is fundamental to computational chemistry and drug formulation research, yet existing benchmarks remain fragmented across datasets, properties, and evaluation protocols, limiting systematic evaluation of model…
Paper v110 Nov
ARE-bench
Research agents achieve increasingly strong benchmark scores, yet these scores do not necessarily translate into reliable delivery in ongoing research projects.
Paper v109 Nov
Paper v109 Nov
Paper v109 Nov
TRIAD
Multimodal retrieval increasingly supports agentic systems, where user requests can involve cross-level relations, combine multiple constraints, and evolve across dialogue turns.
Paper v109 Nov
SWE-Time
Repository-level code-understanding benchmarks have progressively expanded the spatial scope of reasoning, from local code to entire repositories, while largely treating each repository as a static snapshot.
Paper v109 Nov
ShadowBench-MCP
LLM agents increasingly act as routers for tools, APIs, and software-supply-chain workflows, making an execution path - rather than generated text alone - a security-relevant model output.
Paper v109 Nov
Infrathon
In this work, we study whether frontier agent can continually improve real-world AI infrastructure over long horizons.
Paper v109 Nov
Paper v109 Nov
DEPUTYBENCH
Enterprise agents must determine whether ordinary requests are permitted by organizational records and policies while continuing legitimate work.
Paper v109 Nov
ChemInsight
We introduce ChemInsight, a benchmark designed to evaluate large language models (LLMs) on chemistry tasks that require multi-step reasoning and structured chemical understanding.
Paper v109 Nov
AutoBuddy-Bench
As LLM agents become more autonomous, they are increasingly expected to run as always-on assistants on a user's behalf, identifying work to be done and completing it without being asked.
Paper v109 Nov
AgentScript-Eval
Domain-specific languages (DSLs) are widely used in enterprise software, giving a compact interface to complex tasks, from data access to infrastructure automation.