general leaderboards: 248 curated resources
general leaderboards
248 resources · page 4 of 8
langfuse/langfuse
by Langfuse
Langfuse GitHub repo: open-source LLM tracing, evaluation and observability platform.
lechmazur Confabulations
by lechmazur
lechmazur Confabulations GitHub repo: document-based benchmark measuring how often LLMs hallucinate answers in RAG settings.
LIBERO
LIBERO project site: robot manipulation task suites in simulation for evaluating knowledge transfer in lifelong robot learning.
LIBERO-Plus
by Senyu Fei et al. (Tongji University)
LIBERO-Plus paper: a robustness analysis of vision-language-action models using systematically perturbed LIBERO manipulation tasks.
LIBERO-PRO
by Xueyang Zhou et al. (Huazhong University of Science and Technology)
LIBERO-PRO paper testing vision-language-action models under perturbed objects, instructions and environments to check if LIBERO scores generalise.
LiveBench
by LiveBench team (Abacus.AI, NYU, NVIDIA and others)
LiveBench 2.0: regularly refreshed, contamination-limited benchmark of long-horizon real-work tasks.
LLM-jp Eval
by llm-jp
llm-jp-eval GitHub repo: tool for automatically evaluating Japanese LLMs across multiple datasets.
Arena Leaderboard (formerly LMArena)
Arena (formerly LMArena) leaderboards: human-preference rankings across text, webdev, vision, image, video and search arenas.
LongBench
by THUKEG
LongBench GitHub repo from THUDM: bilingual long-context benchmarks LongBench v1 and v2 with datasets and evaluation code.
ManiSkill
by Hillbot and UC San Diego (ManiSkill)
ManiSkill site: GPU-accelerated simulation environments for training and evaluating robot manipulation policies at scale.
MASK (Scale Labs leaderboard)
by Scale AI
Scale Labs MASK leaderboard: benchmark by Scale AI and CAIS measuring honesty separately from accuracy.
MERA-Evaluation/MERA
MERA GitHub repo: Russian-language instruction benchmark with open task definitions and an evaluation harness.
MERA (Alliance in AI)
MERA site in Russian: open benchmark evaluating foundation models on Russian-language tasks, with leaderboards.
MERA leaderboard and ecosystem
MERA portal: Russian-language benchmark leaderboard for foundation models, with methodology and task categories.
MERA leaderboard (RU)
MERA leaderboard in Russian: current results for Text, Industrial, Code and Multi tracks with CSV and JSON export.
MERA
MERA English site: open Russian-language benchmark for foundation models with Text, Reason, Code and Industrial tracks.
Meta-World
by meta-world
Meta-World site: 50 simulated robot manipulation tasks for benchmarking meta-reinforcement and multi-task learning.
MILU
by AI4Bharat
AI4Bharat MILU dataset: multiple-choice questions across subjects in Indian languages for evaluating India-specific knowledge.
Minari
Minari documentation: Python library and dataset standard for offline reinforcement learning research.
MIZAN Persian LLM Leaderboard
by MCINext
MIZAN Persian LLM leaderboard: Persian benchmarks including MT-Bench, IFEval and PerMMLU.
MMLU
MMLU GitHub repo: the original Measuring Massive Multitask Language Understanding test with 57 subjects and evaluation code.
MMLU-Pro
by TIGER Lab
MMLU-Pro GitHub repo: code and data for a harder, ten-option version of MMLU for evaluating language models.
MMLU-ProX
by LiLab
MMLU-ProX dataset: MMLU-Pro translated into many languages for cross-lingual reasoning evaluation.
embeddings-benchmark/mteb
by Massive Text Embedding Benchmark
MTEB GitHub repo: Massive Text Embedding Benchmark code for evaluating embedding models across tasks and languages.
Music Arena
by Music Arena team (Carnegie Mellon University)
Music Arena: arena where users vote between anonymous text-to-music generations to rank models.
NAVSIM (v2)
NAVSIM GitHub repo: pseudo-simulation benchmark for evaluating end-to-end autonomous-driving planners.
NAVSIM v2 navhard leaderboard
by AGC 2025
NAVSIM v2 navhard leaderboard Space: public leaderboard for end-to-end autonomous driving planning.
Nejumi / Japanese LLM Leaderboard
W&B llm-leaderboard GitHub repo: code and methodology for the Nejumi Leaderboard, which evaluates LLMs on Japanese tasks.
Nejumi LLM Leaderboard 4
by Weights & Biases Japan (CoreWeave)
Nejumi LLM Leaderboard 4 report (Weights & Biases Japan, now on CoreWeave Forge): evaluation of LLMs on Japanese tasks.
NoLiMa
by Ali Modarressi et al. (Munich Center for Machine Learning)
NoLiMa paper: a long-context benchmark where the needle must be found without literal word overlap with the question.
nuScenes leaderboards
nuScenes leaderboards: autonomous-driving benchmarks for detection, tracking, prediction and planning.
OALL Hugging Face organization
by Open Arabic LLM Leaderboard
Open Arabic LLM Leaderboard (OALL) Hugging Face organisation: evaluation of Arabic language models.