arxiv arXiv cs.AI · 14h ago · research

Agentic meta-reasoning scales inference through explicit control

from English

The authors introduce agentic meta-reasoning, an inference-time harness that structures execution choices for long-horizon tasks. A controller manages the process by consolidating progress and dispatching work based on a compact run account rather than full history.

On ProgramBench, meta-reasoning achieves 71.5% with GPT-5.5 against 58.0% for Codex, and 67.2% with Opus 4.8 against 65.5% for Claude Code. Across abstract reasoning, multi-domain long-horizon reasoning, and proof generation benchmarks, it gains between 3.6 and 4.2 points over direct control. Artifact-graph analysis shows increased reuse of earlier work and higher coverage of correct solutions.

The results indicate that spending computation on structured control becomes increasingly important as agents scale to longer runs.

Importance 2/3 New harness with differentiators arXiv cs.AI AI agents Research paper

Benchmarks

Benchmark	Model	Score
SWE-Lancer	GPT-5.5	71.5%
SWE-Lancer	Claude Opus 4.8	67.2%

Read original

arxiv arXiv cs.LG · 6d ago · 7 views

GRASP framework improves strategic planning with multi-stage agentic AI

Researchers introduce GRASP, a strategy-aware, multi-stage planning framework designed to generate high-quality natural language executable plans for complex tasks by decoupling the pipeline into specialized modules. The system pre-compiles global macro-guidelines via GenPlan, explores localized strategies through RevPlan, and evaluates trajectories using a multi-criteria discriminator called VerPlan.

arxiv arXiv cs.AI · 6d ago · 6 views

GRASP introduces strategy-aware multi-stage planning to improve LLM reliability

Researchers introduce GRASP, a strategy-aware, multi-stage planning framework designed to generate high-quality natural language executable plans for complex tasks where standard Large Language Models typically degrade in reliability. The system decouples the planning pipeline into specialized modules: GenPlan for global macro-guidelines, RevPlan for exploring localized strategies, and VerPlan for independent trajectory evaluation.

lab Anthropic News · 7d ago · 40 views

Anthropic's Claude discovers novel ART enzyme system in bacteriophages

Anthropic has announced a new life sciences research group dedicated to using Claude to identify uncharacterized protein families and test hypotheses through laboratory experiments. In early results, the AI autonomously discovered a novel enzyme system called array-associated reverse transcriptases (ART) found in bacteriophages.

arxiv arXiv cs.AI · 8d ago · 25 views

AIDE^2 enables recursive self-improvement of AI research agents

Researchers present AIDE^2, a system that implements a loop of recursive self-improvement for an AI research agent by optimizing its own code. The system proposes changes to its internal logic, benchmarks modified versions on AI R&D tasks, and retains only those improvements that perform best on hidden evaluations.

arxiv arXiv cs.AI · 8d ago · 26 views

Growing Harness turns recurring agent control into reusable code via failure-guided training

The authors introduce Growing Harness, a training paradigm that learns an agent's control structure from task feedback rather than relying on standard context-heavy harnesses. By converting recurring control decisions into executable code and reserving LLM calls for semantic reasoning, the method aims to create reusable specialist agents.