GameHorizon-Bench 详情→
GameHorizon Suite introduces GameHorizon-Bench, a benchmark for evaluating gameplay capabilities across horizons with offline and online tracks, covering 21 games and 47 models.
Benchmark 目录
共 2,853 条公开记录，全部通过同一发布门槛（审核通过 · 资源归属已核验 · 已发布）。用下面的筛选缩小范围；代码与数据集状态在各条目内单独标注。
代码
数据集
领域
模态
当前筛选
公开记录
显示前 24 条 / 共 2,853 条BackTrend 详情→
BackTrend is a retrospective benchmark for evaluating scientific weak-signal prediction, containing 25 mature AI/ML topics and 66 human-validated weak signals, with tasks for LLMs, RAG systems, and agentic research systems.
MSI-Bench 详情→
MSI-Bench is a new benchmark for evaluating multi-speaker voice interaction in AI agents, with 1,152 test cases in English and Mandarin, targeting memory, instruction following, and reasoning.
EvoPathBench 详情→
EvoPathBench is a benchmark for process-level evaluation of self-evolving agents, tracking individual capabilities across artifact checkpoints using public trading data and calibrated trajectories.
DUMA-Bench 详情→
DUMA-Bench is a benchmark and evaluation protocol for measuring LLM agent security under dual-control interaction, where both agent and user influence the environment. It extends tau^2-bench with adversarial environments covering eight vulnerability classes and evaluates 14 model
UK-PRBench 详情→
UK-PRBench is a new benchmark for paragraph-level precedent retrieval in UK case law, constructed from UK National Archives judgments. It evaluates state-of-the-art retrieval models and shows that this task remains challenging.
MemCalib 详情→
MemCalib is a benchmark for evaluating how LLM agents use memory in realistic memory-system scenarios, revealing over- and under-use issues. It also introduces an optimization algorithm, MemCalib-RL, but the primary contribution is the benchmark.
IMPLICIT-Bench 详情→
IMPLICIT-Bench is a benchmark for measuring implicit bias in text-to-image models using 5,493 controlled prompt triplets across 11 bias categories, with validation via multi-model agreement, CLIP-based verification, and human evaluation.
MCP-GRANITE 详情→
MCP-GRANITE is an open-source benchmark framework for evaluating MCP-based LLM agents under edge and IoT scenarios, with 81 multi-step scenarios across 9 domains at 4 granularity levels, measuring task completion, tool selection, argument accuracy, latency, and resource usage.
MCPGen 详情→
MCPGen is an executable benchmark for Model Context Protocol workflow development, containing 100 MCP projects across 16 domains, evaluating LLMs on three diagnostic tasks with static analysis, unit/integration tests, and end-to-end execution.
VibeMemBench 详情→
VibeMemBench is a benchmark for evaluating memory systems in coding agents, using 111 coding targets from 90 SWE-rebench V2 repositories with 3,634 history trajectories. It measures whether memory systems improve executable task outcomes compared to baselines.
OmicsBench 详情→
OmicsBench is a new reasoning benchmark for multi-omics sequences with 1,160 expert-validated questions across six tasks, requiring traceable evidence chains and evaluated with instance-specific rubrics.
RiverVLN 详情→
RiverVLN is introduced as the first benchmark for long-horizon USV VLN under continuous riverine motion, with a Unity-ROS closed-loop evaluation protocol and real-world deployment tests.
MuLA-Bench 详情→
MuLA-Bench is a new benchmark with 5,038 open-ended questions over 1,769 in-the-wild recordings across 16 languages, designed to evaluate audio-language models on long-form audio understanding with controlled language-domain and acoustic tracks.
TicTacBench 详情→
TicTacBench is a benchmark with 30 tasks for evaluating coding agents' ability to perform RTL-level timing closure under post-PnR evaluation, using suboptimal RTL designs, timing constraints, and verification.
CraftBench-UE 详情→
CraftBench-UE is an evaluation harness and benchmark with 70 tasks for coding agents in Unreal Engine, using deterministic build, asset, and runtime checks without an LLM judge.
LD-RSVIS 详情→
LD-RSVIS is a new benchmark for robust and general referring surgical video instrument segmentation.
ISA-Bench 详情→
ISA-Bench is a benchmark of programming games with constrained instruction sets, providing full execution stacks for automated evaluation and iterative refinement. It evaluates LLMs' reasoning across unfamiliar instruction set architectures.
MIS-Bench 详情→
MIS-Bench is a new benchmark for evaluating multimodal LLMs on assessing psychotherapeutic interpersonal skills, consisting of 996 annotated videos across 8 dimensions. It reveals gaps in current MLLMs' performance and introduces a fine-tuning method to improve scoring.
SEABED 详情→
SEABED is a new benchmark dataset for evaluating audio reasoning in Southeast Asian speech, comprising 5,404 question-answer pairs across six tasks, and includes evaluation of audio LLMs.
BrainWideBench 详情→
BrainWideBench is a benchmark for evaluating across-animal transfer in multi-region neural recordings, with three task suites for behavior decoding, neural activity prediction, and anatomical organization recovery.
QuranicMMLU 详情→
QuranicMMLU is a benchmark for evaluating generative AI on Quranic Arabic linguistic knowledge, with 980 human-reviewed questions across five linguistic pillars, stratified by cognitive level and verse difficulty, and tested on 12 systems.
RecreationBench 详情→
RecreationWorld introduces RecreationBench, a benchmark of 250 tasks across platforms for hybrid computer-use agents, with reference-grounded programmatic and visual assertions for automatic scoring.
persona-driven IE benchmark 详情→
The paper introduces a new benchmark for personalized information extraction with 292 simulated enterprise users, paired with a persona-generation pipeline, to evaluate LLM-based prompt adaptation methods.