coding agents

2026-10-23

RESEARCH · 1 SOURCE · arXiv cs.AI

RegretBench and BackUpper: measuring and automating backups for coding agents

An arXiv paper (arXiv:2610.28934v1) introduces RegretBench, a 205-task benchmark of real-world "regret" moments where users wished they'd preserved state, and BackUpper, a harness plug-in that checks tool calls and preserves affected state before modifications. Across six frontier models agents complete 86.1% of tasks but back up before regret moments only 15.2%; BackUpper covers all regret actions, restores states with 92.2% average similarity, costs about $0.014 per task, uses 7× fewer tokens than a self-backup baseline, and can match user backup preferences with an 84.5% F1 score.

7.0

2026-10-23

RESEARCH · 1 SOURCE · arXiv cs.AI

GenXR-Bench: first benchmark for coding agents in interactive XR development

The paper (arXiv:2610.29014v1) introduces GenXR-Bench, a benchmark with 112 expert-authored tasks (24 feature implementations, 21 bug repairs, 67 application generations) and held‑out tests combining scripted and adaptive interactions plus programmatic and visual assertions. The authors validated the test suite (96.5% agreement with blinded expert judgments on 124 candidate apps), evaluated 10 models (best configuration achieved 50.9% overall Pass@1), and identify recurring failure modes around asynchronous sensor lifecycles, physical grounding, and embodied interaction semantics.

7.0

2026-09-29

CODING · 1 SOURCE · NVIDIA Developer

NVIDIA publishes lessons from building AI-native TensorRT Model Connect

NVIDIA published a first‑hand account of engineering choices behind TensorRT Model Connect, an open‑source C++ collection of model reference implementations built on TensorRT. The article describes an AI‑native development approach—using coding agents with outcome‑driven tasks, model‑family isolation, reversible changes, and GPU‑backed validation—and notes the project covered 128 model families tested on NVIDIA GB300 as of the public July 29, 2026 release comparison.

6.0

2026-09-18

RESEARCH · 1 SOURCE · arXiv cs.AI

SIFT — Recursive Self-Improvement via Fast Tree-search (arXiv:2609.19526v1)

The paper introduces SIFT, a sample-efficient self-improvement framework for coding agents that uses an LLM-as-a-judge to perform pairwise comparisons of candidate code patches. Judge scores are aggregated with a regularized Bradley–Terry model to guide a lightweight disaggregated tree search, reserving expensive downstream task evaluations for the most promising nodes; the method reportedly outperforms prior tree-search self-evolution approaches on the full Polyglot benchmark while using far fewer CPU hours, wall-clock time, and API cost.

6.0

2026-09-06

RESEARCH · 1 SOURCE · OpenAI

Research acceleration: Inside OpenAI — coding agents and research velocity

OpenAI presents a look inside its use of coding agents and shares early data on agent usage, experiment velocity, task complexity, and how these tools affect research acceleration.

7.0

2026-09-03

COMPANIES · 1 SOURCE · Hugging Face

Hugging Face: 'Give Your Coding Agents a Memory You Own'

Hugging Face published an article titled "Give Your Coding Agents a Memory You Own" about enabling user-controlled memory for coding/AI agents. The source text was not provided, so specific details of the proposal or product are not available here.

5.0