Human Feedback and Eval Paper Explorer

Match reason: Keyword overlap 1/1 across title and protocol fields.

• Therefore, we present Seeded Elimination with Adaptive LLM-as-a-Meta-Judge, a self-improving evaluation protocol for extracting latent ranking signal from saturated benchmarks.
• We evaluate SEAL on multiple saturated benchmarks covering code generation, mathematical reasoning, knowledge-intensive question answering, and tool-use agent task completion.

Match reason: Keyword overlap 1/1 across title and protocol fields.

• Although large language models (LLMs) have significantly advanced text generation, stories generated by LLMs still diverge from human-authored works regarding complex narrative structure and human-aligned preferences.
• We find existing reward models struggle to select human-preferred stories, with the best model achieving only 66.3\% accuracy.

Match reason: Keyword overlap 1/1 across title and protocol fields.

• We introduce IslamicMMLU, a benchmark of 10,013 multiple-choice questions spanning three tracks: Quran (2,013 questions), Hadith (4,000 questions), and Fiqh (jurisprudence, 4,000 questions).
• The benchmark is used to create the IslamicMMLU public leaderboard for evaluating LLMs, and we initially evaluate 26 LLMs, where their averaged accuracy across the three tracks varied between 39.8\% to 93.8\% (by Gemini 3 Flash).

Match reason: Keyword overlap 1/1 across title and protocol fields.

• To address these challenges, we introduce SDiaReward, an end-to-end multi-turn reward model trained on SDiaReward-Dataset, a novel collection of episode-level preference pairs explicitly targeting these gaps.
• Experiments demonstrate that SDiaReward achieves state-of-the-art pairwise preference accuracy, significantly outperforming general-purpose audio LLMs.

Match reason: Keyword overlap 1/1 across title and protocol fields.

• In the random-error setting, models strongly prefer correct completions in paired evaluation: 83.1% accuracy at balanced data and 67.0% even when correct rules appear in only 10% of the corpus.
• Replacing random errors with a coherent but mathematically incorrect rule system largely eliminates the preference (near-chance accuracy).

Match reason: Keyword overlap 1/1 across title and protocol fields.

• The models were developed through a four-stage training pipeline: continued pre-training on Portuguese and Brazilian legal corpora, long-context extension to 128K tokens, supervised fine-tuning on instruction data spanning chat, code, legal…
• We evaluate the models on six benchmark categories: conversational capabilities in Brazilian Portuguese, knowledge of Brazilian legislation, long-context understanding, instruction following, standardized exams, and agentic capabilities…

Match reason: Keyword overlap 1/1 across title and protocol fields.

• On this unified benchmark, we evaluate four approaches: (i) encoder-based classification fine-tuning, (ii) zero- and few-shot prompting, (iii) instruction tuning and Retrieval-Augmented Generation (RAG), and (iv) Supervised Fine-Tuning…

Match reason: Keyword overlap 1/1 across title and protocol fields.

• While prior research emphasizes the role of on-policy data in mitigating forgetting, we uncover--and validate both theoretically and empirically--an overlooked yet critical mechanism: the implicit regularization inherent in Direct…

Match reason: Keyword overlap 1/1 across title and protocol fields.

• In this work, we propose ProAgent, an end-to-end proactive agent system that harnesses on-demand sensory contexts to provide in-the-wild assistance.
• Results demonstrate that ProAgent achieves up to 27.7% higher proactive prediction accuracy and 20.5% lower false detection than state-of-the-art baselines.

Match reason: Keyword overlap 1/1 across title and protocol fields.

• Finally, we demonstrate efficiency enhancements across three downstream tasks: preference data synthesis, Pareto frontier approximation and optimization, and distillation of aligned behaviors for intervention-free inference.

Match reason: Keyword overlap 1/1 across title and protocol fields.

• We evaluate MLLM verifiers across web navigation, computer use, and robotics, spanning 13+ models, 28+ designs, and thousands of trajectories from diverse agents.
• Our methods yield more human-aligned verifiers, improving failure detection by 25pp and accuracy by 14pp.

Match reason: Keyword overlap 1/1 across title and protocol fields.

• We introduce Vibe Code Bench, a benchmark of 100 web application specifications (50 public validation, 50 held-out test) with 964 browser-based workflows comprising 10,131 substeps, evaluated against deployed applications by an autonomous…
• We identify self-testing during generation as a strong performance predictor (Pearson r=0.72), and show through a completed human alignment study that evaluator selection materially affects outcomes (31.8-93.6% pairwise step-level…

Match reason: Keyword overlap 1/1 across title and protocol fields.

• Reward models are central to aligning large language models (LLMs) with human preferences.
• Yet most approaches rely on pointwise reward estimates that overlook the epistemic uncertainty in reward models arising from limited human feedback.