The authors create the first large-scale dataset and taxonomy of failure modes in multi-agent LLM systems to explain their limited performance gains.
super hub Mixed citations
Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
Mixed citation behavior. Most common role is background (47%).
209
Pith papers citing it
Background
47% of classified citations
abstract
Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge.
hub tools
citation-role summary
background 15
method 10
dataset 4
baseline 1
citation-polarity summary
claims ledger
- abstract Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-be
authors
co-cited works
representative citing papers
ORPO performs preference alignment during supervised fine-tuning via a monolithic odds ratio penalty, allowing 7B models to outperform larger state-of-the-art models on alignment benchmarks.
LongBench is the first bilingual multi-task benchmark for long context understanding in LLMs, containing 21 datasets in 6 categories with average lengths of 6711 words (English) and 13386 characters (Chinese).
Gradient and greedy search over token suffixes produces universal, transferable adversarial prompts that elicit objectionable outputs from aligned models including black-box commercial systems.
Analysis of 500k ChatGPT logs shows over one-third of conversations generate fiction, dominated by power users with repetitive and niche patterns.
Low-resource safety failures are action failures because the harmfulness representation transfers but the decision calibration does not; this is fixed by recalibrating a high-resource gate with 1-4 target-language examples.
RWGBench is a citation-centric benchmark for related work generation built from 40k CS papers and a 100-paper test set, with multi-dimensional metrics that better match human expert judgment than standard similarity scores.
Introduces (ε,q,t,A)-behavioral indistinguishability and shows via Qwen/Llama experiments that LoRA distillation boosts semantic similarity but leaves detectable behavioral differences under adversarial evaluation.
A Behavioral Specification interpretive layer improves representational accuracy for AI personalization by compressing user data into patterns, outperforming raw corpora and commercial memory systems on held-out behavioral predictions across 14 autobiographical corpora while reducing context cost.
OR-Space is a benchmark for LLM agents performing full-lifecycle optimization tasks across Build, Revise, and Explain modes in executable multi-artifact workspaces.
Self-evolving rubric with anti-gaming fitness reveals that objective capability scaling fails to transfer to subjective LLM behaviors, with advice-restraint as the universal lowest dimension that can regress.
LLMs show severe staleness after training cutoffs and recency bias on historical German statutes; RAG with version filtering mitigates both better than web search.
A multi-agent pipeline iteratively refines topology optimization outputs to match natural language preferences for branched structures, achieving 60% success rate across replicates in cantilever and phone-stand tasks.
Introduces the stochastic-deterministic boundary (SDB) as a load-bearing primitive for LLM agent runtimes and provides a five-step methodology plus catalog of six patterns adapted from distributed systems.
DecisionBench supplies a fixed task suite, model pool, delegation interface, and multi-axis metrics to evaluate emergent delegation, showing similar quality across awareness conditions but 15-31 point headroom under perfect delegation.
SCICONVBENCH is a new benchmark evaluating LLMs on multi-turn disambiguation and inconsistency resolution for task formulation in computational science, with frontier models reaching only 52.7% success on fluid mechanics disambiguation cases.
CBEA with LCV bounds evidence sets and validates commitments before response generation, achieving zero failures in scoped tests at 0.49-0.60 availability versus near-zero for baselines.
Belief Engine is a configurable belief-update mechanism for multi-agent LLM systems that uses structured argument extraction and log-odds stance updates to make evidence-grounded deliberation inspectable and controllable.
Test-Time Hinting trains a hint generator to prepend contextual guidance to VLM prompts, improving accuracy on natural-image VQA benchmarks with generalization to unseen tasks and models.
Presents a likelihood-based benchmark for equation-suffix prediction in technical papers with controls to detect shortcut vulnerabilities in model forecasts.
LLM multi-agent systems on lattices show bias-driven order-disorder crossovers instead of true phase transitions, with extracted effective couplings and fields serving as model-specific fingerprints.
SlimSpec replaces the standard LM-head in draft models with a low-rank version to deliver 4-5x faster speculative decoding while preserving full vocabulary and competitive acceptance rates.
ProactBench measures LLM conversational proactivity in three phases using 198 multi-agent dialogues and finds recovery behavior hard to predict from existing benchmarks.
LLMs routinely produce unsupported causal stories for personal sensing anomalies, and richer evidence or constrained prompts do not reliably eliminate this epistemic overreach.
citing papers explorer
-
Why Do Multi-Agent LLM Systems Fail?
The authors create the first large-scale dataset and taxonomy of failure modes in multi-agent LLM systems to explain their limited performance gains.
-
ORPO: Monolithic Preference Optimization without Reference Model
ORPO performs preference alignment during supervised fine-tuning via a monolithic odds ratio penalty, allowing 7B models to outperform larger state-of-the-art models on alignment benchmarks.
-
LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding
LongBench is the first bilingual multi-task benchmark for long context understanding in LLMs, containing 21 datasets in 6 categories with average lengths of 6711 words (English) and 13386 characters (Chinese).
-
Universal and Transferable Adversarial Attacks on Aligned Language Models
Gradient and greedy search over token suffixes produces universal, transferable adversarial prompts that elicit objectionable outputs from aligned models including black-box commercial systems.
-
AI Fiction in the Wild
Analysis of 500k ChatGPT logs shows over one-third of conversations generate fiction, dominated by power users with repetitive and niche patterns.
-
Low-Resource Safety Failures Are Action Failures, Not Representation Failures
Low-resource safety failures are action failures because the harmfulness representation transfers but the decision calibration does not; this is fixed by recalibrating a high-resource gate with 1-4 target-language examples.
-
RWGBench: Evaluating Scholarly Positioning in Related Work Generation
RWGBench is a citation-centric benchmark for related work generation built from 40k CS papers and a 100-paper test set, with multi-dimensional metrics that better match human expert judgment than standard similarity scores.
-
Bounded Behavioral Indistinguishability for Black-Box LLM Distillation
Introduces (ε,q,t,A)-behavioral indistinguishability and shows via Qwen/Llama experiments that LoRA distillation boosts semantic similarity but leaves detectable behavioral differences under adversarial evaluation.
-
Beyond Recall: Behavioral Specification as an Interpretive Layer for AI Personalization
A Behavioral Specification interpretive layer improves representational accuracy for AI personalization by compressing user data into patterns, outperforming raw corpora and commercial memory systems on held-out behavioral predictions across 14 autobiographical corpora while reducing context cost.
-
OR-Space: A Full-Lifecycle Workspace Benchmark for Industrial Optimization Agents
OR-Space is a benchmark for LLM agents performing full-lifecycle optimization tasks across Build, Revise, and Explain modes in executable multi-artifact workspaces.
-
Does Capability Transfer to Subjective Behavior -- and Would Our Instruments Tell Us? A Self-Evolving, Trust-by-Construction Evaluation Paradigm
Self-evolving rubric with anti-gaming fitness reveals that objective capability scaling fails to transfer to subjective LLM behaviors, with advice-restraint as the universal lowest dimension that can regress.
-
Asking For An Old Friend: Diagnosing and Mitigating Temporal Failure Modes in LLM-based Statutory Question Answering
LLMs show severe staleness after training cutoffs and recency bias on historical German statutes; RAG with version filtering mitigates both better than web search.
-
TO-Agents: A Multi-Agent AI Pipeline for Preference-Guided Topology Optimization
A multi-agent pipeline iteratively refines topology optimization outputs to match natural language preferences for branched structures, achieving 60% success rate across replicates in cantilever and phone-stand tasks.
-
A Methodology for Selecting and Composing Runtime Architecture Patterns for Production LLM Agents
Introduces the stochastic-deterministic boundary (SDB) as a load-bearing primitive for LLM agent runtimes and provides a five-step methodology plus catalog of six patterns adapted from distributed systems.
-
DecisionBench: A Benchmark for Emergent Delegation in Long-Horizon Agentic Workflows
DecisionBench supplies a fixed task suite, model pool, delegation interface, and multi-axis metrics to evaluate emergent delegation, showing similar quality across awareness conditions but 15-31 point headroom under perfect delegation.
-
SCICONVBENCH: Benchmarking LLMs on Multi-Turn Clarification for Task Formulation in Computational Science
SCICONVBENCH is a new benchmark evaluating LLMs on multi-turn disambiguation and inconsistency resolution for task formulation in computational science, with frontier models reaching only 52.7% success on fluid mechanics disambiguation cases.
-
Recall Isn't Enough: Bounding Commitments in Personalized Language Systems
CBEA with LCV bounds evidence sets and validates commitments before response generation, achieving zero failures in scoped tests at 0.49-0.60 availability versus near-zero for baselines.
-
Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM Deliberation
Belief Engine is a configurable belief-update mechanism for multi-agent LLM systems that uses structured argument extraction and log-odds stance updates to make evidence-grounded deliberation inspectable and controllable.
-
Test-Time Hinting for Black-Box Vision-Language Models
Test-Time Hinting trains a hint generator to prepend contextual guidance to VLM prompts, improving accuracy on natural-image VQA benchmarks with generalization to unseen tasks and models.
-
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.
-
Collective Alignment in LLM Multi-Agent Systems: Disentangling Bias from Cooperation via Statistical Physics
LLM multi-agent systems on lattices show bias-driven order-disorder crossovers instead of true phase transitions, with extracted effective couplings and fields serving as model-specific fingerprints.
-
SlimSpec: Low-Rank Draft LM-Head for Accelerated Speculative Decoding
SlimSpec replaces the standard LM-head in draft models with a low-rank version to deliver 4-5x faster speculative decoding while preserving full vocabulary and competitive acceptance rates.
-
ProactBench: Beyond What The User Asked For
ProactBench measures LLM conversational proactivity in three phases using 198 multi-agent dialogues and finds recovery behavior hard to predict from existing benchmarks.
-
Causal Stories from Sensor Traces: Auditing Epistemic Overreach in LLM-Generated Personal Sensing Explanations
LLMs routinely produce unsupported causal stories for personal sensing anomalies, and richer evidence or constrained prompts do not reliably eliminate this epistemic overreach.
-
GameGen-Verifier: Parallel Keypoint-Based Verification for LLM-Generated Games via Runtime State Injection
GameGen-Verifier decomposes game specifications into keypoints, injects runtime states for targeted checks, and achieves 92.2% accuracy on 100 games while running up to 16.6x faster than agent-based baselines.
-
Instruction Tuning Changes How Upstream State Conditions Late Readout: A Cross-Patching Diagnostic
Instruction tuning makes late-layer computation depend more on the model's own post-trained upstream state than on base-model upstream state, producing a consistent +1.68 logit interaction effect across five model families.
-
EditPropBench: Measuring Factual Edit Propagation in Scientific Manuscripts
EditPropBench evaluates LLM editors on propagating factual edits to dependent claims in synthetic scientific manuscripts, showing that even the strongest systems miss roughly 30% of required updates on hard cases.
-
The Partial Testimony of Logs: Evaluation of Language Model Generation under Confounded Model Choice
An identification theorem shows that a randomized experiment and simulator together recover causal model values from confounded logs, with logs used only afterward to reduce estimation error.
-
Subliminal Steering: Stronger Encoding of Hidden Signals
Subliminal steering transfers complex behavioral biases and the underlying steering vector through fine-tuning on innocuous data, achieving higher precision than prior prompt-based methods.
-
Agentic Witnessing: Pragmatic and Scalable TEE-Enabled Privacy-Preserving Auditing
Agentic Witnessing enables privacy-preserving auditing of semantic properties in private data by running an LLM auditor in a TEE that answers binary queries and produces cryptographic transcripts of its reasoning.
-
AsmRAG: LLM-Driven Malware Detection by Retrieving Functionally Similar Assembly Code
AsmRAG detects malware at 96% F1 and attributes families at 95% F1 by retrieving functionally similar assembly code via LLM embeddings and density-weighted anchor selection, remaining robust to metamorphic obfuscation.
-
HiPO: Hierarchical Preference Optimization for Adaptive Reasoning in LLMs
HiPO improves LLM reasoning performance by optimizing preferences separately on response segments rather than entire outputs.
-
More Thinking, More Bias: Length-Driven Position Bias in Reasoning Models
Position bias scales positively with reasoning trajectory length in CoT models, shown by partial correlations and truncation interventions across multiple benchmarks and model scales.
-
Don't Start What You Can't Finish: A Counterfactual Audit of Support-State Triage in LLM Agents
LLM agents overcommit on non-complete tasks at 41.7% unless given explicit support-state categories, which raise typed deferral accuracy to 91.7%.
-
Evaluating Tool-Using Language Agents: Judge Reliability, Propagation Cascades, and Runtime Mitigation in AgentProp-Bench
AgentProp-Bench shows substring judging agrees with humans at kappa=0.049, LLM ensemble at 0.432, bad-parameter injection propagates with ~0.62 probability, rejection and recovery are independent, and a runtime fix cuts hallucinations 23pp on GPT-4o-mini but not Gemini-2.0-Flash.
-
The cognitive companion: a lightweight parallel monitoring architecture for detecting and recovering from reasoning degradation in LLM agents
A parallel Cognitive Companion architecture reduces repetition in LLM agents by 52-62% on loop-prone tasks using LLM monitoring with 11% overhead or zero-overhead probes on hidden states, with benefits depending on task type.
-
CompliBench: Benchmarking LLM Judges for Compliance Violation Detection in Dialogue Systems
CompliBench uses simulation and adversarial flaw injection to create labeled dialogue data showing that top proprietary LLMs perform poorly at spotting guideline violations while fine-tuned smaller models outperform them and generalize to new domains.
-
Vision-Language Foundation Models for Comprehensive Automated Pavement Condition Assessment
Instruction-tuned vision-language model PaveGPT, trained on a large unified pavement dataset, achieves substantial gains over general models in comprehensive, standard-compliant pavement condition assessment.
-
IoT-Brain: Grounding LLMs for Semantic-Spatial Sensor Scheduling
IoT-Brain uses a neuro-symbolic Spatial Trajectory Graph to ground LLMs for verifiable semantic-spatial sensor scheduling, achieving 37.6% higher task success with lower resource use on a campus-scale benchmark.
-
An Agentic Evaluation Architecture for Historical Bias Detection in Educational Textbooks
An agentic architecture with multimodal screening, a five-agent jury, meta-synthesis, and source attribution protocol detects biases in Romanian history textbooks more accurately than zero-shot baselines, achieving 83.3% acceptable excerpts and human preference in 64.8% of blind comparisons.
-
FrontierFinance: A Long-Horizon Computer-Use Benchmark of Real-World Financial Tasks
FrontierFinance benchmark shows human financial experts outperform state-of-the-art LLMs by achieving higher scores and more client-ready outputs on realistic long-horizon tasks.
-
LatentAudit: Real-Time White-Box Faithfulness Monitoring for Retrieval-Augmented Generation with Verifiable Deployment
LatentAudit monitors RAG faithfulness in real time via Mahalanobis distance on residual-stream activations, reaching 0.942 AUROC on PubMedQA with 0.77 ms overhead and supporting Groth16 verification.
-
Synthetic Data Generation for Training Diversified Commonsense Reasoning Models
A two-stage synthetic data generation method creates the CommonSyn dataset, allowing LLMs fine-tuned on it to produce more diverse and higher-quality commonsense responses than vanilla or human-data-trained models.
-
CapTrack: Multifaceted Evaluation of Forgetting in LLM Post-Training
CapTrack shows post-training causes drift beyond facts, with instruction fine-tuning producing stronger behavioral changes than preference optimization across model families.
-
Bayesian Preference Learning for Test-Time Steerable Reward Models
ICRM casts reward modeling as amortized variational inference over a latent preference probability with a Beta prior, enabling test-time adaptation to unseen preferences and improving benchmark performance.
-
Results-Actionability Gap: Understanding How Practitioners Evaluate LLM Products in the Wild
Qualitative study of 19 practitioners reveals ten LLM product evaluation practices and introduces the results-actionability gap as a key barrier to turning findings into improvements.
-
Omni-R1: Towards the Unified Generative Paradigm for Multimodal Reasoning
Omni-R1 unifies multimodal reasoning by generating intermediate images during the process in a SFT-plus-RL framework, with an Omni-R1-Zero variant that matches or exceeds it using only text data.
-
VLegal-Bench: Cognitively Grounded Benchmark for Vietnamese Legal Reasoning of Large Language Models
VLegal-Bench supplies 10,450 expert-validated samples for evaluating LLMs on Vietnamese legal questions, retrieval, multi-step reasoning, and scenario solving.
-
Evalet: Evaluating Large Language Models through Functional Fragmentation
Evalet applies functional fragmentation to deliver fragment-level qualitative analysis of LLM evaluations, with a user study showing 48% more misalignment detections than holistic scoring.
-
LLMs Judging LLMs: A Simplex Perspective
A simplex geometry framework shows when LLM-as-judge rankings are identifiable, provides a basis for preferring binary over multi-level scoring, and uses Bayesian priors to model epistemic uncertainty about judges.