Hung-Chun Hsu, Po-Jen Ko, Che-Cheng Wu, Li-Yang Chang, Chuan-Ju Wang · Aug 12, 2026 · Citations: 0
Llm As JudgeAutomatic MetricsGeneral
Predictive-distribution entropy is a strong answer-selection rule in retrieval-augmented generation (RAG) for question answering: across five QA benchmarks, selecting the answer a frozen respondent LLM produces with the lowest answer-token…
LODESTAR uses reinforcement learning (GRPO) to train, once and offline, a polarizer -- a short fixed natural-language string inserted into the respondent's prompt and never into its weights, directing entropy so that entropy-based answer…
Large language models are increasingly deployed in settings that require simultaneous adherence to multiple explicit constraints - reasoning structure, safety boundaries, output schemas.
We introduce Constraint Saturation Evaluation (CSE), a procedurally generated benchmark that systematically varies the number of simultaneous constraints (k), with every constraint scored by a deterministic, rule-based verifier and zero…
Large language model (LLM) agents integrated with external tools are vulnerable to indirect prompt injections embedded in environmental states.
To bridge this gap, we propose **ToolHazard**, a scalable adversarial environment synthesis framework that reduces human engineering and supports expansion with additional seed domains and compute.
Ofir Ben Shoham, Shrutendra Harsola, Vignesh Subrahmaniam, Shravan Mohan, Yakov Gazman, Oded Vainas · Aug 12, 2026 · Citations: 0
Rubric RatingLlm As JudgeGeneral
Our reward is an LLM-as-a-judge rubric that scores each recommendation across multiple binary dimensions of advice quality, augmented with a safety gate for harm prevention.
Since LLM-based evaluation alone cannot confirm whether improvements reflect genuine business value rather than adaptation to the judge, we complement it with a judge-independent audit based on a standard doubly-robust Conditional Average…
Long-form song generation models continue to improve in duration, structural integrity, and acoustic complexity, making reliable aesthetic rewards increasingly important for aligning these models with human preferences.
On the out-of-domain Music Arena benchmark with 733 preference pairs, it achieves the highest accuracy of 71.35%.
As users increasingly turn to Large Language Models (LLMs) for information and advice on political matters, particularly during election periods, the political preferences expressed by these systems have become a matter of public interest.
We demonstrate the framework through an Italian case study, providing a systematic analysis of LLM-generated political evaluations on italian parties and leaders.
Nimet Beyza Bozdag, Emre Can Acikgoz, Gokhan Tur, Dilek Hakkani-Tür · Aug 12, 2026 · Citations: 0
Automatic MetricsGeneral
As LLMs increasingly debate, advise, and think collaboratively with humans and each other, resistance to harmful persuasion becomes a core requirement for reliable behavior.
We formalize this threat as adversarial persuasion and introduce an adversarial reinforcement learning framework that trains persuader agents to change a target model's answer in a single interaction.
Vu Duc Anh, Nhat M. Hoang, Do Xuan Long, Cong-Duy Nguyen, Ponhvoan Srey, Luu Anh Tuan · Aug 12, 2026 · Citations: 0
Pairwise PreferenceGeneral
The first stage strengthens step-level reasoning via step-level preference optimization, while the second stage explicitly trains models to self-verify and self-correct.
Comprehensive in-domain and out-of-domain evaluations across multiple LLMs demonstrate that SFS-DPO and SFS-DPO-R consistently outperform prior step-level training baselines.
For LLM agents, however, the unit of observation is an interactive trajectory, where the model can ask clarifying questions, call tools, update state, and make intermediate decisions whose errors propagate to the final outcome.
Haokai Zhao, Yunze Xiao, Weihao Xuan, Flora Salim, Benjamin Tag, Aditya Joshi · Aug 12, 2026 · Citations: 0
Pairwise PreferenceGeneral
Group alignment adapts a language model to a demographic group to produce responses that reflect the group's opinions, values, and preferences.
However, existing group alignment methods and evaluations focus only on how closely the model matches the group's opinions, overlooking the induced change in sycophantic behaviour.
On the full GPQA Diamond benchmark (198 graduate-level science questions), majority voting reduces per-problem accuracy on a majority of problems for two instruction-tuned models from different families: 56.6% of problems for Qwen2.5-7B and…
Rima Mittal, Ankit Gubrani, Satyanarayana Kakollu · Aug 11, 2026 · Citations: 0
Pairwise PreferenceGeneral
Through controlled experiments on two GPU families spanning the memory-bound to compute-bound regimes (A10G, ridge \approx 117 FLOP/byte; A100, ridge \approx 183 FLOP/byte), we demonstrate: (1) the inference-optimal vocabulary shifts 16x…
Natural language user preferences provide an interpretable interface for LLM personalization.
To this end, we propose AlignXada, a training-free meta-learning framework that induces reusable textual refinement policies for adapting universal preference summaries to task-specific ones.
Long-horizon LLM agents increasingly rely on external execution support to maintain state, track progress, recover from failures, and reuse experience across extended interactions.
Yet existing harnesses and their use are often tailored to environments and controlled through prompts, heuristics, or system-specific rules, making agent and harness coordination difficult to jointly optimize.
Alicia Guerra, Yibo Hu · Aug 5, 2026 · Citations: 0
Critique EditGeneral
Open-ended revisions require a different measurement strategy because answer quality is graded, latent, and judged imperfectly.
We introduce an experimental protocol implemented across a pooled main peer-condition corpus and separately constructed decomposition corpora, allowing us to separate ordinary re-answering, candidate-content exposure, a bundled…
Zhilin Wang, Han Song, Runzhe Zhan, Jusen Du, Jiacheng Chen, Tianle Li · Jul 2, 2026 · Citations: 0
Simulation EnvGeneral
Autonomous agents are increasingly expected to improve executable policies through feedback, yet existing evaluations often collapse this process into a final score or confound it with open-ended software-engineering progress.
We introduce Autonomous Policy Evolution, a controlled evaluation setting in which a harness-model agent repeatedly edits an executable policy system under a fixed interaction budget.
Manuel Alonso-Carracedo, Ruben Fernandez-Boullon, Pedro Celard, Francisco J. Rodriguez-Martinez, Lorena Otero-Cerdeira · Jul 2, 2026 · Citations: 0
Rubric RatingAutomatic MetricsGeneral
Gemini~3.0 Pro with rubric-guided prompting achieved the highest human-AI agreement (ICC(3,1) = 0.888, MAE = 0.10, Bland-Altman bias = -0.014).
These results show that question complexity is a reliable predictor of the difficulty LLMs face in grading accurately, and they establish a principled, taxonomy-based framework for determining which questions are suitable for AI-assisted…
Memory for a long-horizon LLM agent is a contract about what each future decision is allowed to see.
A public online benchmark of frontier LLMs on the same game reports zero wins at the lowest difficulty across five configurations, and the developer-reported human win rate at the same difficulty is 16%; the task is hard but not saturated.
