From Dev.to Community
Tech Articles
Curated developer articles, tutorials, and guides — auto-updated hourly
Peremptory6d ago • 2 min read
Frontier Models Fail at Research-Level Reasoning
New benchmark shows frontier LLMs can only recover research ideas from bibliographies 3-15% of the t...
#benchmarks#research#aidevelopment
0 0
PeremptoryAug 21, 2026 • 2 min read
Frontier Models Hit a Wall on Research Thinking
A new benchmark shows frontier LLMs recover research paper ideas from bibliographies alone at just 3...
#research#benchmarks#aidevelopment
0 0
Artyom KornilovAug 21, 2026 • 13 min read
Predicting Code Duplication Detection Performance: Model Specifications and Benchmarks Fall Short
Introduction Code duplication detection is a cornerstone of modern software development,...
#codeduplication#embeddingmodels#benchmarks#performance
0 0
Pradeep KumarAug 24, 2026 • 12 min read
The Best Open Source LLMs in August 2026: A Developer's Honest Guide
Open-weight models now match closed frontiers on independent benchmarks. The question is no longer '...
#aimodels#opensourcemodels#benchmarks#opensourcemodelsaugust2026
1 0
Breach ProtocolAug 22, 2026 • 4 min read
The ARC-AGI-3 record going around is the wrong number and the wrong system
The top ARC-AGI-3 entry on ARC Prize's public leaderboard is an NVIDIA-labelled agent scoring 85.1% ...
#benchmarks#agents#arcagi#evaluation
0 0
Breach Protocol5d ago • 4 min read
A new benchmark of 1,140 real agent failures finds the best method identifies the decisive wrong step 13 percent of the time
LongRCA Bench collects 1,140 genuinely failed agent runs averaging 145 steps each, with human labels...
#agents#benchmarks#debugging#evaluation
0 0
Breach ProtocolAug 21, 2026 • 4 min read
Video models look right 92 percent of the time and do the task 38 percent
A new benchmark scores AI video generation on two separate axes and finds the best model reaches 91....
#videogeneration#benchmarks#evaluation#multimodal
0 0
Breach ProtocolAug 23, 2026 • 4 min read
Coding agents ace the public test and stumble on the hidden one
A new benchmark of 119 real scientific software tasks keeps its grading tests private, and the top a...
#benchmarks#codingagents#science#evaluation
0 0
Breach Protocol5d ago • 4 min read
Microsoft's AutoSaddler treats the agent harness as code to be patched, and gains about ten points on three benchmarks
Microsoft researchers built a system that reads an agent's failure traces, writes structured patches...
#agents#harness#microsoft#benchmarks
0 0