Daily Archive
Updated from current HFEPX corpus (Feb 27, 2026). 11 papers are grouped in this daily page. Common evaluation modes: Automatic Metrics, Human Eval. Common annotation unit: Scalar. Frequently cited benchmark: WebShop. Common metric signal: accuracy. Newest paper in this set is from Feb 12, 2026.
Updated from current HFEPX corpus (Feb 27, 2026). This page covers 11 papers centered on HFEPX Daily Archive: 2026-02-12. Common evaluation modes include Automatic Metrics, Human Eval, with benchmark emphasis on WebShop, Zoombench. Use the anchored takeaways below to compare protocol choices and identify papers with stronger evidence depth.
Evaluation emphasis: Automatic Metrics and Human Eval appear frequently in this slice.
Evidence: propella-1: Multi-Property Document Annotation for LLM Data Curation at Scale , Think like a Scientist: Physics-guided LLM Agent for Equation Discovery , "Sorry, I Didn't Catch That": How Speech Models Miss What Matters Most
Benchmark concentration: WebShop, Zoombench helps control cross-paper variance.
Evidence: Think like a Scientist: Physics-guided LLM Agent for Equation Discovery , "Sorry, I Didn't Catch That": How Speech Models Miss What Matters Most , Scaling Model and Data for Multilingual Machine Translation with Open Large Language Models
Metric concentration: accuracy, agreement is repeatedly reported in this group.
Evidence: "Sorry, I Didn't Catch That": How Speech Models Miss What Matters Most , Scaling Model and Data for Multilingual Machine Translation with Open Large Language Models , Who is the richest club in the championship? Detecting and Rewriting Underspecified Questions Improve QA Performance
Pair this hub with llm_as_judge pages to benchmark automated-vs-human evaluation tradeoffs.
Evidence: Scaling Model and Data for Multilingual Machine Translation with Open Large Language Models , Who is the richest club in the championship? Detecting and Rewriting Underspecified Questions Improve QA Performance , Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perception
Stratify by benchmark (WebShop vs Zoombench) before comparing methods.
Evidence: Who is the richest club in the championship? Detecting and Rewriting Underspecified Questions Improve QA Performance , Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perception , TSR: Trajectory-Search Rollouts for Multi-Turn RL of LLM Agents
Papers with explicit human feedback is visible in approximately 9.1% of papers in this set.
Evidence: Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perception , TSR: Trajectory-Search Rollouts for Multi-Turn RL of LLM Agents , Curriculum Learning and Pseudo-Labeling Improve the Generalization of Multi-Label Arabic Dialect Identification Models
Papers with explicit human feedback
Coverage is a replication risk (9.1% vs 45% target).
Papers reporting quality controls
Coverage is a replication risk (0% vs 30% target).
Papers naming benchmarks/datasets
Coverage is a replication risk (18.2% vs 35% target).
Papers naming evaluation metrics
Coverage is strong (54.5% vs 35% target).
Papers with known rater population
Coverage is a replication risk (0% vs 35% target).
Papers with known annotation unit
Coverage is a replication risk (18.2% vs 35% target).
1. propella-1: Multi-Property Document Annotation for LLM Data Curation at Scale
Start with this anchor paper for scope and protocol framing. Covers Human Eval.
2. Think like a Scientist: Physics-guided LLM Agent for Equation Discovery
Covers Automatic Metrics.
3. "Sorry, I Didn't Catch That": How Speech Models Miss What Matters Most
Covers Automatic Metrics.
4. Scaling Model and Data for Multilingual Machine Translation with Open Large Language Models
Covers Automatic Metrics.
Covers Automatic Metrics.
6. Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perception
Covers Automatic Metrics.
7. TSR: Trajectory-Search Rollouts for Multi-Turn RL of LLM Agents
Covers Simulation Env.
Covers Automatic Metrics.
human_eval vs automatic_metrics
both=0, left_only=1, right_only=9
0 papers use both Human Eval and Automatic Metrics.
automatic_metrics vs simulation_env
both=0, left_only=9, right_only=1
0 papers use both Automatic Metrics and Simulation Env.
human_eval vs simulation_env
both=0, left_only=1, right_only=1
0 papers use both Human Eval and Simulation Env.
Maximilian Idahl, Benedikt Droste, Björn Plüster, Jan Philipp Harries · Feb 12, 2026
We introduce propella-1, a family of small multilingual LLMs (0.6B, 1.7B, 4B parameters) that annotate text documents across 18 properties organized into six categories: core content, classification, quality and value, audience and purpose,
Jianke Yang, Ohm Venkatachalam, Mohammad Kianezhad, Sharvaree Vadgama, Rose Yu · Feb 12, 2026
We introduce KeplerAgent, an agentic framework that explicitly follows this scientific reasoning process.
Kaitlyn Zhou, Martijn Bartelds, Federico Bianchi, James Zou · Feb 12, 2026
Despite speech recognition systems achieving low word error rates on standard benchmarks, they often fail on short, high-stakes utterances in real-world deployments.
Yuzhe Shang, Pengzhi Gao, Wei Liu, Jian Luan, Jinsong Su · Feb 12, 2026
Open large language models (LLMs) have demonstrated improving multilingual capabilities in recent years.
Yunchong Huang, Gianni Barlacchi, Sandro Pezzelle · Feb 12, 2026
Large language models (LLMs) perform well on well-posed questions, yet standard question-answering (QA) benchmarks remain far from solved.
Lai Wei, Liangbo He, Jun Lan, Lingzhong Dong, Yutong Cai · Feb 12, 2026
To address this, we propose Region-to-Image Distillation, which transforms zooming from an inference-time tool into a training-time primitive, thereby internalizing the benefits of agentic zooming into a single forward pass of an MLLM.
Aladin Djuhera, Swanand Ravindra Kadhe, Farhan Ahmed, Heiko Ludwig, Holger Boche · Feb 12, 2026
Advances in large language models (LLMs) are driving a shift toward using reinforcement learning (RL) to train agents from iterative, multi-turn interactions across tasks.
Ali Mekky, Mohamed El Zeftawy, Lara Hassan, Amr Keleg, Preslav Nakov · Feb 12, 2026
Being modeled as a single-label classification task for a long time, recent work has argued that Arabic Dialect Identification (ADI) should be framed as a multi-label classification task.
Maomao Li, Zhen Li, Kaipeng Zhang, Guosheng Yin, Zhifeng Li · Feb 12, 2026
Third, we train OmniCustom on our constructed large-scale, high-quality audio-visual human dataset.
Sri Durga Sai Sowmya Kadali, Evangelos E. Papalexakis · Feb 12, 2026
Jailbreaking large language models (LLMs) has emerged as a critical security challenge with the widespread deployment of conversational AI systems.
Jayadev Billa · Feb 12, 2026
Using ALME, a benchmark of 57,602 controlled audio-text conflict stimuli across 8 languages, we find that Gemini 2.0 Flash exhibits 16.6% text dominance under audio-text conflict versus 1.6% under text-text conflict with identical reliabili
