A new benchmark with cognitive traps shows frontier deep research agents achieve only 13-16% acceptance on expert consulting tasks under combined verifier and rubric criteria.
hub Mixed citations
Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators
Mixed citation behavior. Most common role is background (33%).
80
Pith papers citing it
Background
33% of classified citations
abstract
LLM-based auto-annotators have become a key component of the LLM development process due to their cost-effectiveness and scalability compared to human-based evaluation. However, these auto-annotators can introduce biases that are hard to remove. Even simple, known confounders such as preference for longer outputs remain in existing automated evaluation metrics. We propose a simple regression analysis approach for controlling biases in auto-evaluations. As a real case study, we focus on reducing the length bias of AlpacaEval, a fast and affordable benchmark for instruction-tuned LLMs that uses LLMs to estimate response quality. Despite being highly correlated with human preferences, AlpacaEval is known to favor models that generate longer outputs. We introduce a length-controlled AlpacaEval that aims to answer the counterfactual question: "What would the preference be if the model's and baseline's output had the same length?" To achieve this, we first fit a generalized linear model to predict the biased auto-annotator's preferences based on the mediators we want to control for (length difference) and other relevant features. We then obtain length-controlled preferences by predicting preferences while conditioning the GLM with a zero difference in lengths. Length-controlling not only improves the robustness of the metric to manipulations in model verbosity, but we also find that it increases the Spearman correlation with LMSYS Chatbot Arena from 0.94 to 0.98.
hub tools
citation-role summary
background 3
dataset 3
baseline 1
method 1
other 1
citation-polarity summary
representative citing papers
LiveBench is a contamination-limited LLM benchmark with auto-scored challenging tasks from recent sources across math, coding, reasoning and more, where top models score below 70%.
CV-Arena is a new 12K-pair benchmark for instruction-guided real-image editing with 16 task types, CogRetriever curation, and Active Elo mixed human-AI evaluation that finds gaps in 21 models and presents CV-Agent.
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.
DRATS derives a minimax objective from a feasibility formulation of MTRL to adaptively sample tasks with the largest return gaps, leading to better worst-task performance on MetaWorld benchmarks.
Abstract-CoT lets models reason with short discrete latent token sequences from a reserved vocabulary, using warm-up training and RL to match verbal CoT performance with up to 11.6x fewer tokens.
Continuous adversarial training in the embedding space produces a robust generalization bound for linear transformers that decreases with perturbation radius, tied to singular values of the embedding matrix, and motivates a new regularizer that improves real LLM jailbreak robustness-utility tradeoff
SalesLLM provides an automatic evaluation framework for LLM sales dialogues that correlates 0.98 with human experts and shows top models approaching human performance while weaker ones lag.
TiCo enables spoken dialogue models to follow explicit time constraints in generated responses using Spoken Time Markers and reinforcement learning with verifiable rewards, cutting duration error by 2.7x over its backbone.
CyclicJudge uses round-robin judge-to-scenario assignment to recover the panel-mean score exactly while using the same number of judge calls as single-judge evaluation.
Info-Gain Sampler improves MDM decoding by using bidirectional information gain to reduce cumulative uncertainty, outperforming greedy samplers on reasoning accuracy and creative writing tasks.
Top-W applies Wasserstein-regularized truncation on token-embedding geometry to create a closed-form optimal crop for LLM sampling that outperforms prior methods by up to 33.7% on GSM8K, GPQA, AlpacaEval, and MT-Bench.
VideoP2R separates perception and reasoning in a process-aware RFT pipeline with a new CoT dataset and PA-GRPO rewards, reaching SOTA on six of seven video benchmarks.
A well-tuned kNN router matches or exceeds state-of-the-art learned routers on new standardized benchmarks spanning instruction, QA, reasoning, and the first multi-modal visual routing dataset, due to locality of model performance in embedding space.
Magpie synthesizes 300K high-quality alignment instructions from Llama-3-Instruct via auto-regressive prompting on partial templates, enabling fine-tuned models to match official instruct performance on AlpacaEval, ArenaHard, and WildBench.
DeepSeek-V2 delivers top-tier open-source LLM performance using only 21B active parameters by compressing the KV cache 93.3% and cutting training costs 42.5% via MLA and DeepSeekMoE.
CogWM is a new LLM user model for evaluating social influence by predicting and tracking cognitive state evolution in dialogues, trained on 150k samples and shown to differentiate AI agents effectively.
Proposes SCSuff metric for evaluating LLM explanation sufficiency via model-generated alternative inputs, showing explanations are typically insufficient and predictable from hidden states.
Curating concise data for VLMs induces brevity, delivering 35x lower Cost-of-Pass at near-identical accuracy and higher matched-length accuracy than uncurated baselines.
FOXGLOVE dataset of 2340 comments shows LLMs and instructors align on feedback goals and positions but diverge on sentence selection, with LLMs using more complex language and fewer questions and higher quality ratings driven by comment length.
SEAL revives saturated benchmarks via adaptive LLM meta-judging in elimination matches, matching full pairwise accuracy with roughly half the calls across code, math, QA, and agent tasks.
Persona prompting trades expertise depth for reduced clarity in LLM answers and works best on advisory questions in medicine and psychology.
RUBRIC-ARROW is an alternating rubric generator and judge framework that uses probability-based scoring and pairwise preferences to improve pointwise reward modeling accuracy for LLM post-training in non-verifiable domains.
AdaDPO uses self-adaptive stop-gradient coefficients to balance preferred and dispreferred gradients in DPO, achieving higher AlpacaEval 2 win rates than standard DPO on Llama-3-8B-Instruct.
citing papers explorer
-
Does Math Reasoning Improve General LLM Capabilities? Understanding Transferability of LLM Reasoning
Math reasoning gains in LLMs rarely transfer to general domains; RL tuning generalizes while SFT causes forgetting and representation drift.
-
XtraGPT: Context-Aware and Controllable Academic Paper Revision via Human-AI Collaboration
XtraGPT is a suite of 1.5B-14B parameter open-source LLMs fine-tuned on 140,000 revision pairs from 7,000 top-tier papers to support controllable, context-aware academic paper editing.
-
The Differences Between Direct Alignment Algorithms are a Blur
A controlled unification of direct alignment algorithms shows the ranking objective (pairwise vs pointwise) drives alignment quality more than the scalar score optimized.
-
DataComp-LM: In search of the next generation of training sets for language models
DCLM-Baseline dataset lets a 7B model reach 64% 5-shot MMLU accuracy after 2.6T tokens, beating prior open-data models by 6.6 points on MMLU with 40% less compute.
-
DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence
An open-source MoE code model matches GPT-4 Turbo on coding and math benchmarks while expanding to 338 languages and 128K context length.
-
Mixture-of-Agents Enhances Large Language Model Capabilities
A layered Mixture-of-Agents system combining multiple LLMs achieves state-of-the-art results on AlpacaEval 2.0 (65.1%), MT-Bench, and FLASK, outperforming GPT-4 Omni.
-
Can MLLMs Critique Like Humans? Evaluating Open-Ended Aesthetic Reasoning in Multimodal Large Language Models
MLLMs generate verbose, comprehensive, and repetitive aesthetic critiques unlike selective human ones, and reference-based metrics fail to detect this because they capture model house style instead of image-specific content.
-
CARE-RL: Capability-Aware Reinforcement Learning for Mitigating Cross-Domain Conflicts
CARE-RL combines PA-GRM for task-adaptive rewards on open-ended tasks and DACSP for modulating RL updates using historical capability directions, reporting higher total average scores than baselines on Qwen models.
-
A Pilot Study on Curator-Guided Multilingual Art Description for Blind and Low-Vision Audiences with Small Vision-Language Models
Pilot evaluation of language-specific versus multilingual LoRA adapters on Qwen2.5-VL-3B for curator-guided BLV art descriptions in three languages.
-
TRACE: Toulmin-based Reasoning Assessment through Constructive Elements for LLM CoT Evaluation
TRACE is a new metric for assessing LLM CoT reasoning structure via Toulmin and Flavell frameworks, showing r=0.74 correlation with accuracy on 26.3K samples and utility as an RL reward.
-
A Fixed-Budget, Cluster-Aware Standard for LLM-as-a-Judge Evaluation: A Multi-Hop RAG Stress Test
Proposes a minimum measurement standard for LLM-as-a-judge in multi-hop RAG that fixes budgets and requires cluster-aware inference, showing it alters which baseline comparisons remain significant.
-
CroCo: Cross-Lingual Contrastive Preference Tuning on Self-Generations
CroCo applies English-reward-ranked self-generations for contrastive preference tuning that improves two LLMs on structured and open-ended tasks across 14 languages without language-specific annotations.
-
LoCar: Localization-Aware Evaluation of In-Vehicle Assistants through Fine-Grained Sociolinguistic Control
LoCar is a localization-aware evaluation framework for in-vehicle assistants that identifies unstable Korean honorific control and weaker performance on strategic metrics like clarification and proactivity in current LLMs.
-
Re-Triggering Safeguards within LLMs for Jailbreak Detection
Embedding disruption re-triggers LLM internal safeguards to detect jailbreak prompts more effectively than standalone defenses.
-
Data-dependent Exploration for Online Reinforcement Learning from Human Feedback
DEPO constructs uncertainty bonuses from historical data for exploration in online RLHF and provides a data-dependent regret bound that adapts to task hardness.
-
A Validated Prompt Bank for Malicious Code Generation: Separating Executable Weapons from Security Knowledge in 1,554 Consensus-Labeled Prompts
The paper releases a 1,554-prompt consensus-labeled bank separating executable malicious code requests from security knowledge requests, validated by five-model majority labeling with Fleiss' kappa of 0.876.
-
LLM-as-a-Judge for Human-AI Co-Creation: A Reliability-Aware Evaluation Framework for Coding
LLM judges for human-AI coding co-creation show moderate performance (ROC-AUC 0.59) and low agreement, with co-creation success concentrating early in interactions.
-
MedConclusion: A Benchmark for Biomedical Conclusion Generation from Structured Abstracts
MedConclusion is a 5.7M-instance benchmark dataset for generating biomedical conclusions from structured PubMed abstracts, with LLM evaluations showing conclusion writing differs from summarization and that judge choice affects scores.
-
What Is Preference Optimization Doing, and Why?
Gradient analysis and ablations show DPO and PPO have different target directions and component roles in preference optimization for LLMs.
-
GIFT: Group-Relative Implicit Fine-Tuning Integrates GRPO with DPO and UNA
GIFT matches the optimal policy of GRPO using an endogenous prompt-dependent KL coefficient derived via z-score standardization of implicit rewards.
-
Proximal Supervised Fine-Tuning
PSFT modifies supervised fine-tuning by incorporating trust-region ideas from RL to constrain policy changes, yielding better out-of-domain generalization in math and human-value tasks without entropy collapse.
-
Enhancing Speech Large Language Models through Reinforced Behavior Alignment
Reinforced Behavior Alignment (RBA) uses self-synthesized data from a teacher LLM and reinforcement learning to close the instruction-following gap in SpeechLMs, outperforming distillation and reaching SOTA on spoken QA and speech-to-text translation benchmarks.
-
InfiGFusion: Graph-on-Logits Distillation via Efficient Gromov-Wasserstein for Model Fusion
InfiGFusion introduces graph-on-logits distillation with an O(n log n) Gromov-Wasserstein approximation to fuse LLMs by modeling token co-activations, reporting gains over baselines on 11 benchmarks.
-
Multiple Choice Questions: Reasoning Makes Large Language Models (LLMs) More Self-Confident, Especially When They are Wrong
Reasoning before answering MCQs increases LLM confidence more for incorrect answers and degrades calibration on a 57-subject benchmark across seven models.
-
WildFeedback: Aligning LLMs With In-situ User Interactions And Feedback
WildFeedback extracts preference pairs from in-situ user feedback in LLM conversations to fine-tune models for better alignment with real user preferences.
-
Reward-Free Code Alignment from Pretrained or Fine-Tuned LLM: Unpacking the Trade-offs for Code Generation
Empirical study on five LLMs finds pretrained-to-aligned paths yield bigger gains over baseline than finetuned-to-aligned paths, though absolute accuracy remains lower for pretrained starts.
-
LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods
A survey that organizes LLMs-as-judges research into functionality, methodology, applications, meta-evaluation, and limitations.
- Judging the Judges: A Systematic Evaluation of Bias Mitigation Strategies in LLM-as-a-Judge Pipelines
- IatroBench: Pre-Registered Evidence of Iatrogenic Harm from AI Safety Measures
- Reinforcement Learning from Human Feedback