mini-SWE-agent
PulseAugur coverage of mini-SWE-agent — every cluster mentioning mini-SWE-agent across labs, papers, and developer communities, ranked by signal.
5 day(s) with sentiment data
-
AI agent performance hinges on harness, not just model, study finds
A recent experiment involving 27,280 agent runs and 83 billion tokens revealed that the surrounding harness configuration, rather than the underlying language model, significantly impacts performance and cost. Across 22…
-
Coding agent harnesses show minimal score impact, but costs vary widely · 1 source tracked
A recent study published on arXiv and covered by VentureBeat investigated the impact of different coding agent harnesses on LLM performance. Researchers from Shandong University found that when using the same underlying…
-
Coding agent harnesses increase costs without improving performance, study finds
A recent study indicates that employing various coding agent harnesses, such as Claude Code, OpenCode, or mini-SWE-agent, does not significantly improve their success rates on coding tasks. However, these harnesses can …
-
New SWE-Verifier model boosts coding agent efficiency and accuracy
Researchers have developed SWE-Verifier, a lightweight generative model designed to improve the efficiency and decision-making of coding agents. This model selectively triggers candidate sampling and verification proces…
-
New ReuF2F benchmark tests AI coding agents on formal math problems
Researchers have introduced ReuF2F, a new benchmark designed to evaluate AI coding agents on formalized undergraduate research problems. This benchmark consists of 244 open problems drawn from Research Experiences for U…
-
New benchmark Parsimony measures AI code generation efficiency
Parsimony is a new open-source benchmark designed to evaluate AI models based on the amount of code they add, rather than just their success rate. Developed to address the long-term maintainability of codebases, Parsimo…
-
Mini-SWE-agent shows promise in debugging benchmarks, using fewer tokens than GPT-5.6
A user conducted a benchmark comparing the mini-swe-agent with GPT-5.6 "Sol" for debugging tasks. The mini-swe-agent, particularly when utilizing a "bash + linear history" setup, demonstrated a significantly higher pass…
-
New tuning method boosts LLM coding agent performance
Researchers have developed a new method called probe-and-refine tuning to improve the performance of large language model (LLM) coding agents. This technique focuses on enhancing the guidance files that direct agents to…
-
New benchmark reveals AI agents struggle with research nuance
A new benchmark series called AARR has been introduced to evaluate the research capabilities of advanced AI agents. The first iteration, AARRI-Bench, tests agents on tasks requiring professionalism, thoroughness, and nu…