MathConstraint generates scalable, automatically verifiable combinatorial problems where LLMs achieve 18.5-66.9% accuracy without tools but roughly double that with solver access.
hub Canonical reference
LiveBench: A Challenging, Contamination-Limited LLM Benchmark
Canonical reference. 78% of citing Pith papers cite this work as background.
58
Pith papers citing it
18
external citations · Pith
Background
78% of classified citations
abstract
Test set contamination, wherein test data from a benchmark ends up in a newer model's training set, is a well-documented obstacle for fair LLM evaluation and can quickly render benchmarks obsolete. To mitigate this, many recent benchmarks crowdsource new prompts and evaluations from human or LLM judges; however, these can introduce significant biases, and break down when scoring hard questions. In this work, we introduce a new benchmark for LLMs designed to be resistant to both test set contamination and the pitfalls of LLM judging and human crowdsourcing. We release LiveBench, the first benchmark that (1) contains frequently-updated questions from recent information sources, (2) scores answers automatically according to objective ground-truth values, and (3) contains a wide variety of challenging tasks, spanning math, coding, reasoning, language, instruction following, and data analysis. To achieve this, LiveBench contains questions that are based on recently-released math competitions, arXiv papers, news articles, and datasets, and it contains harder, contamination-limited versions of tasks from previous benchmarks such as Big-Bench Hard, AMPS, and IFEval. We evaluate many prominent closed-source models, as well as dozens of open-source models ranging from 0.5B to 405B in size. LiveBench is difficult, with top models achieving below 70% accuracy. We release all questions, code, and model answers. Questions are added and updated on a monthly basis, and we release new tasks and harder versions of tasks over time so that LiveBench can distinguish between the capabilities of LLMs as they improve in the future. We welcome community engagement and collaboration for expanding the benchmark tasks and models.
hub tools
citation-role summary
background 6
dataset 3
citation-polarity summary
representative citing papers
TabArena launches a dynamic, updatable benchmarking system for tabular ML that shows boosted trees remain competitive, deep learning matches them under larger budgets with ensembling, foundation models excel on small data, and cross-model ensembles advance SOTA while flagging validation overfitting.
A meta-benchmarking framework organizes 452 LLM benchmarks into 41 O*NET Generalized Work Activities and 38 BIAN domains, using discrimination-coverage-recency weights to scale K-factors in an Elo tournament for comparable financial-services scores.
AlgoBench creates traceable variants of competitive programming problems via constraint shifts that invalidate original algorithms, paired with complexity metrics that reveal LLMs often produce functionally correct but asymptotically unsuitable solutions.
Introduces the Power Systems Agent Benchmark with 41 task families across eight power engineering areas for executable evaluation of AI agents using deterministic feasibility checks.
A new code-writing data analysis benchmark shows human experts outperforming a frontier LLM on average with lower performance variance.
The paper introduces Uni-E, a unified energy for DLMs that accounts for model capacity, dependency and invariance, can be computed exactly, and corrects distribution shifts from dependency and invariance.
CoEval generates task-specific benchmarks by rotating models through teacher, student, and judge roles, then weights questions by discriminative power and judges by panel consensus to recover accurate model rankings without labels.
LiveBrowseComp shows search agents rely on intrinsic knowledge on standard benchmarks, with scores dropping 25-40 points and closed-book accuracy below 2% on questions about facts from the prior 90 days.
Frontier LLMs reason well about past science but are near chance at judging whether specific advances will happen, systematically late on when, and overconfident — even with extra pre-cutoff knowledge.
Presents a likelihood-based benchmark for equation-suffix prediction in technical papers with controls to detect shortcut vulnerabilities in model forecasts.
Re²Math is a new benchmark that evaluates AI models on retrieving and verifying the applicability of theorems from math literature to advance steps in partial proofs, accepting any sufficient theorem while controlling for leakage.
MathArena evaluates over 50 LLMs on 162 fresh competition problems across seven contests, detects contamination in AIME 2024, and reports top models scoring below 40 percent on IMO 2025 proof tasks.
PRIMETIME generator reveals that LLM datetime parsing and arithmetic primitives are individually unreliable but fully learnable via fine-tuning, enabling frontier-level accuracy on event planning with small LoRA models.
UnifiedReward is the first unified reward model that jointly assesses multimodal understanding and generation to provide better preference signals for aligning vision models via DPO.
EduArt is a new benchmark of 871 educational questions that reveals multimodal LLMs perform near ceiling on multiple-choice art history items but drop sharply on open completion and error identification tasks.
MMGist filters 23,250 items from 18 benchmarks down to 7,262 using three-stage pipeline, preserving model rankings (Spearman ρ=0.98) while cutting items 69% and raising discrimination 78%.
MetaPS trains models via simulation rollouts to select from programmatic strategy libraries for market agents, yielding better performance than fixed or direct LLM baselines across model sizes.
ReKey introduces a live benchmark protocol that regenerates visual keys in images to produce contamination-resilient VQA evaluations, showing 9.5-18.8 point higher scores on original items across eight VLMs.
AGDO improves dLLM reasoning performance by determining denoising order and emphasizing tokens based on attention-derived dependencies rather than random masking.
A graph-based MIS prompt selection method on embedding similarity graphs yields reduced benchmark subsets with highly consistent LLM rankings (Kendall's W ≥ 0.90 in 99.2% of cases) and 25-48% size reduction at higher thresholds.
TRACER presents a semantic-aware framework and the first benchmark for fine-grained code contamination detection across three levels of overlap, reporting F1 scores of 0.91-0.92 and large gains over prior methods.
SynAE is a multi-metric framework that evaluates how well synthetic benchmarks replicate real data characteristics for multi-turn tool-calling agent testing.
First unified survey formalizing Pretraining Data Exposure across exposure levels and reviewing attack, defense, and contamination methods for LLMs.
citing papers explorer
-
MathConstraint: Automated Generation of Verified Combinatorial Reasoning Instances for LLMs
MathConstraint generates scalable, automatically verifiable combinatorial problems where LLMs achieve 18.5-66.9% accuracy without tools but roughly double that with solver access.
-
TabArena: A Living Benchmark for Machine Learning on Tabular Data
TabArena launches a dynamic, updatable benchmarking system for tabular ML that shows boosted trees remain competitive, deep learning matches them under larger budgets with ensembling, foundation models excel on small data, and cross-model ensembles advance SOTA while flagging validation overfitting.
-
Meta-Benchmarks for Financial-Services LLM Evaluation
A meta-benchmarking framework organizes 452 LLM benchmarks into 41 O*NET Generalized Work Activities and 38 BIAN domains, using discrimination-coverage-recency weights to scale K-factors in an Elo tournament for comparable financial-services scores.
-
AlgoBench: Benchmarking Algorithmic Adaptation in Code Generation
AlgoBench creates traceable variants of competitive programming problems via constraint shifts that invalidate original algorithms, paired with complexity metrics that reveal LLMs often produce functionally correct but asymptotically unsuitable solutions.
-
Power Systems Agent Benchmark: Executable Evaluation of AI Agents in Electric Power Engineering
Introduces the Power Systems Agent Benchmark with 41 task families across eight power engineering areas for executable evaluation of AI agents using deterministic feasibility checks.
-
Flaws in the LLM Automation Narrative
A new code-writing data analysis benchmark shows human experts outperforming a frontier LLM on average with lower performance variance.
-
Unified Energy for Invariant and Independent Decoding in Diffusion Language Models
The paper introduces Uni-E, a unified energy for DLMs that accounts for model capacity, dependency and invariance, can be computed exactly, and corrects distribution shifts from dependency and invariance.
-
CoEval: Ranking Language Models for Custom Tasks Without Labeled Data or Trustworthy Benchmarks
CoEval generates task-specific benchmarks by rotating models through teacher, student, and judge roles, then weights questions by discriminative power and judges by panel consensus to recover accurate model rankings without labels.
-
LiveBrowseComp: Are Search Agents Searching, or Just Verifying What They Already Know?
LiveBrowseComp shows search agents rely on intrinsic knowledge on standard benchmarks, with scores dropping 25-40 points and closed-book accuracy below 2% on questions about facts from the prior 90 days.
-
Scientific reasoning does not reliably translate into scientific forecasting in frontier AI
Frontier LLMs reason well about past science but are near chance at judging whether specific advances will happen, systematically late on when, and overconfident — even with extra pre-cutoff knowledge.
-
Likelihood scoring for continuations of mathematical text: a self-supervised benchmark with tests for shortcut vulnerabilities
Presents a likelihood-based benchmark for equation-suffix prediction in technical papers with controls to detect shortcut vulnerabilities in model forecasts.
-
Re$^2$Math: Benchmarking Theorem Retrieval in Research-Level Mathematics
Re²Math is a new benchmark that evaluates AI models on retrieving and verifying the applicability of theorems from math literature to advance steps in partial proofs, accepting any sufficient theorem while controlling for leakage.
-
MathArena: Evaluating LLMs on Uncontaminated Math Competitions
MathArena evaluates over 50 LLMs on 162 fresh competition problems across seven contests, detects contamination in AIME 2024, and reports top models scoring below 40 percent on IMO 2025 proof tasks.
-
PRIMETIME : Limits of LLMs in Temporal Primitives
PRIMETIME generator reveals that LLM datetime parsing and arithmetic primitives are individually unreliable but fully learnable via fine-tuning, enabling frontier-level accuracy on event planning with small LoRA models.
-
Unified Reward Model for Multimodal Understanding and Generation
UnifiedReward is the first unified reward model that jointly assesses multimodal understanding and generation to provide better preference signals for aligning vision models via DPO.
-
EduArt: An educational-level benchmark for evaluating art history knowledge in large language models
EduArt is a new benchmark of 871 educational questions that reveals multimodal LLMs perform near ceiling on multiple-choice art history items but drop sharply on open completion and error identification tasks.
-
MMGist: A Comprehensive Multimodal Benchmark for 2027
MMGist filters 23,250 items from 18 benchmarks down to 7,262 using three-stage pipeline, preserving model rankings (Spearman ρ=0.98) while cutting items 69% and raising discrimination 78%.
-
MetaPS: Adaptive Programmatic Strategy Selection for Market Agents
MetaPS trains models via simulation rollouts to select from programmatic strategy libraries for market agents, yielding better performance than fixed or direct LLM baselines across model sizes.
-
REKEY: Metadata-Grounded Visual-Key Regeneration for Contamination-Resilient VQA Evaluation
ReKey introduces a live benchmark protocol that regenerates visual keys in images to produce contamination-resilient VQA evaluations, showing 9.5-18.8 point higher scores on original items across eight VLMs.
-
Beyond Fully Random Masking: Attention-Guided Denoising and Optimization for Diffusion Language Models
AGDO improves dLLM reasoning performance by determining denoising order and emphasizing tokens based on attention-derived dependencies rather than random masking.
-
Consistent and Distinctive: LLM Benchmark Efficiency via Maximum Independent Set Prompt Selection on Similarity Graphs
A graph-based MIS prompt selection method on embedding similarity graphs yields reduced benchmark subsets with highly consistent LLM rankings (Kendall's W ≥ 0.90 in 99.2% of cases) and 25-48% size reduction at higher thresholds.
-
TRACER: A Semantic-Aware Framework for Fine-Grained Contamination Detection in Code LLMs
TRACER presents a semantic-aware framework and the first benchmark for fine-grained code contamination detection across three levels of overlap, reporting F1 scores of 0.91-0.92 and large gains over prior methods.
-
SynAE: A Framework for Measuring the Quality of Synthetic Data for Tool-Calling Agent Evaluations
SynAE is a multi-metric framework that evaluates how well synthetic benchmarks replicate real data characteristics for multi-turn tool-calling agent testing.
-
Pretraining Data Exposure in Large Language Models: A Survey of Membership Inference, Data Contamination, and Security Implications
First unified survey formalizing Pretraining Data Exposure across exposure levels and reviewing attack, defense, and contamination methods for LLMs.
-
ShapeCodeBench: A Renewable Benchmark for Perception-to-Program Reconstruction of Synthetic Shape Scenes
ShapeCodeBench introduces a renewable benchmark for perception-to-program reconstruction of synthetic shapes, with evaluations showing low exact-match performance from current models and heuristics.
-
Formalizing Latent Thoughts: Four Axioms of Thought Representation in LLMs
Four axioms (Causality, Minimality, Separability, Stability) are formalized for latent thought representations; audits of open LLMs on 23 tasks show none satisfy all four and representations add little beyond input embeddings.
-
Decision-aware User Simulation Agent for Evaluating Conversational Recommender Systems
Hesitator is a theory-grounded simulator that separates utility-based item selection from overload-aware commitment decisions to reduce unrealistic high acceptance rates in conversational recommender evaluations.
-
You Don't Need Public Tests to Generate Correct Code
DryRUN lets LLMs create their own test inputs and run internal simulations for self-correcting code generation, matching the performance of test-dependent methods like CodeSIM on LiveCodeBench without public tests or external signals.
-
LLMs for Qualitative Data Analysis Fail on Security-specificComments in Human Experiments
LLMs improve with detailed code descriptions but remain insufficient to replace human annotators for security-specific qualitative coding.
-
Babbling Suppression: Making LLMs Greener One Token at a Time
Babbling Suppression stops LLM code generation upon test passage to reduce token output and energy consumption by up to 65% across Python and Java benchmarks.
-
Fighting AI with AI: AI-Agent Augmented DNS Blocking of LLM Services during Student Evaluations
AI-Sinkhole uses AI classification with quantized LLMs and Pi-Hole DNS blocking to dynamically prevent access to LLM services during student evaluations, reporting F1 scores above 0.83.
-
Kimi Linear: An Expressive, Efficient Attention Architecture
Kimi Linear hybridizes linear attention with a new KDA module to beat full attention on tasks while slashing KV cache by 75% and speeding decoding up to 6x.
-
SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?
SWE-Bench Pro is a new benchmark with 1,865 long-horizon tasks from 41 repositories designed to evaluate AI agents on realistic enterprise-level software engineering problems beyond prior benchmarks.
-
EMERGE: A Benchmark for Updating Knowledge Graphs with Emerging Textual Knowledge
EMERGE is a benchmark dataset of 233K Wikipedia passages paired with 1.45 million Wikidata edit operations across seven yearly snapshots from 2019 to 2025 for evaluating knowledge graph updates from emerging text.
-
Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource
MoE models with activation rates in an optimal region outperform dense LLMs of identical total parameter count, training compute, and data budget, with the optimal region consistent across scales.
-
Process Reinforcement through Implicit Rewards
PRIME enables online process reward model updates in LLM RL using implicit rewards from rollouts and outcome labels, yielding 15.1% average gains on reasoning benchmarks and surpassing a stronger instruct model with 10% of the data.
-
Qwen2.5-1M Technical Report
Qwen2.5-1M models reach 1M token context with improved long-context performance, no short-context loss, and 3-7x prefill speedup via open inference optimizations.
-
DOPD: Dual On-policy Distillation
DOPD is an advantage-aware dual distillation method that dynamically assigns token supervision from either privileged teacher or student to transfer capability while mitigating non-replicable information asymmetry in on-policy distillation.
-
When Does Intrinsic Self-Correction Help? A Task-Sensitive Analysis
Intrinsic self-correction produces gains on tasks whose structure supports explicit verification, reasoning revisits, or second opinions, but not as a uniform method.
-
Sakana Fugu Technical Report
Sakana Fugu trains LLM orchestrators using fine-tuning, evolutionary algorithms, and RL to build query-adaptive multi-agent scaffolds, claiming SOTA results on benchmarks including SWE-Bench Pro and GPQA-Diamond.
-
The Emergence of Autonomous Penetration Capabilities in Large Language Model-Powered AI Systems
A tiered server benchmark with 300 targets shows current LLMs achieve autonomous penetration success rates of 10.7-69.3% using only general cybersecurity tools and no target-specific knowledge.
-
Continual LLM Upcycling: A Predictor-Gated Bank-Wise Sparsity Training Recipe for Dense-to-Sparse LLMs
Continual training recipe upcycles dense Qwen2.5-8B LLM to 4x channel-sparse model via predictor-gated bank-wise sparsity in SwiGLU FFN with a single-layer repair for long-context failure on RULER-CWE.
-
Characterize Then Distill: Mechanistic Reasoning in Large Output Spaces
Reasoning in large output spaces proceeds via shortlisting then fine-grained reasoning; this characterization enables a mechanistic distillation strategy that outperforms standard distillation.
-
CodeGolf Bench: A Multi-Language Benchmark for Evaluating Concise Code Generation Capabilities of Large Language Models
CodeGolf Bench is a dynamic benchmark for LLM concise code generation in 60 languages, showing reasoning models reach 70.97% average human percentile on Python and C++ tasks while non-reasoning models lag.
-
PTCG-Bench: Can LLM Agents Master Pok\'emon Trading Card Game?
PTCG-Bench shows LLM agents reach non-trivial PTCG performance but struggle with sustained self-evolution and remain sensitive to harness design.
-
Benchmarking EngGPT2-16B-A3B against Comparable Italian and International Open-source LLMs
EngGPT2MoE-16B-A3B matches or exceeds other Italian open-source LLMs on most international benchmarks while remaining competitive on ITALIC, though it trails some top international models.
-
Language models fail at extended rule following
LLMs fail at extended counting of repeated characters due to finite internal states, with abrupt errors persisting across model scales and inference methods.
-
A Large Language Model Based Pipeline for Review of Systems Entity Recognition from Clinical Notes
LLM pipeline with novel attribution algorithm extracts ROS entities, negation status, and body systems from 24 clinical notes at up to 0.952 F1 using open-source models.
-
Qwen3 Technical Report
Pith review generated a malformed one-line summary.
-
SCAN: Structured Capability Assessment and Navigation for LLMs
SCAN is a framework for fine-grained LLM capability assessment via automatic taxonomy construction from queries, query synthesis for coverage, visualization tools, and a PC2-enhanced LLM-as-a-judge method, applied to 21 models showing intra-family variations.