livecodebench

Tiny Model, Big Logic: Diversity-Driven Optimization Elicits Large-Model Reasoning Ability in VibeThinker-1.5B

[NeurIPS 2025] AGI-Elo: How Far Are We From Mastering A Task?

Benchmark methodology, task sets, and evaluation results for RA²R

Interactive benchmark for evaluating LLMs on clarifying ambiguous code requirements. Paper: arXiv:2607.00711

Resample or reroute after a weak-verifier stop? Pre-registered measurements of recoverable stopping debt on MBPP+, a two-sided action-support gate on BigCodeBench that fails closed, a LiveCodeBench observability ladder, and the exchangeable-actions reference showing realized-maximum gaps carry no selector signal. Artifacts for arXiv:2607.08665v3.

A reproducible LiveCodeBench evaluation scaffold for controlled studies of code-reasoning SFT on Qwen2.5-1.5B.

Self-Improving Agent for Code Generation

LoRA fine-tuning evaluation pipeline for code generation models on LiveCodeBench benchmark

“U.S. Provisional Patent Application No. 63/978,753, filed February 9, 2026” Aleph Generalizable Intelligence Convergence Engine (AGICE) - Evidence-Native, Policy-Governed Reasoning with Rollback and Geometrical Multi-Dimensional Transformation (GMDT) via the Complex Operator (−i)

AI 大模型第三方权威排行榜聚合：LMArena、Artificial Analysis、LiveCodeBench、SuperCLUE、HF Open LLM。无后端纯静态站，GitHub Actions 每日自动抓取更新，含升降箭头/性价比象限/历史趋势。 | Independent AI model leaderboard aggregator - zero-backend static site, auto-updated daily.

NeoSmith Maestro — frontier-level coding accuracy from a bouquet of small models at ~1/30th the cost. Technical report + reproducible evidence (LiveCodeBench v6 92.2%, SWE-bench Pro 74.2/88.9).

Self-hosted LLM evaluation lab: MMLU-Pro / BFCL / IFEval / LiveCodeBench / HumanEval+ runs on Tesla V100 and RTX 3090, llama.cpp serving (incl. Volta sm_70), patched API benchmark clients

Executable-code evaluation and verifier-guided inference, with reproducible benchmarks and separate model-weight retention audits.