HelpSteer3: Human-Annotated Feedback and Edit Data to Empower Inference-Time Scaling in Open-Ended General-Domain Tasks (2025.acl-long)
Copied to clipboard
Zhilin Wang, Jiaqi Zeng, Olivier Delalleau, Daniel Egert, Ellie Evans, Hoo-Chang Shin, Felipe Soares, Yi Dong, Oleksii Kuchaiev
| Challenge: | Inference-Time Scaling is critical to the success of recent models such as OpenAI o1 and DeepSeek R1 . however, many techniques require tasks to have answers that can be verified . |
| Approach: | They use data to train dedicated Feedback and Edit Models capable of inference-time scaling for open-ended tasks. |
| Outcome: | The proposed model can reach SoTA performance on Arena Hard at 92.7 as of 5 Mar 2025. |
Similar Papers
s1: Simple test-time scaling (2025.emnlp-main)
Copied to clipboard
Niklas Muennighoff, Zitong Yang, Weijia Shi, Xiang Lisa Li, Li Fei-Fei, Hannaneh Hajishirzi, Luke Zettlemoyer, Percy Liang, Emmanuel Candes, Tatsunori Hashimoto
| Challenge: | OpenAI’s o1 model showed this capability but did not publicly share its methodology, leading to many replication efforts. |
| Approach: | They curate a small dataset s1K with 1,000 reasoning questions based on three criteria we validate through ablations: difficulty, diversity, and quality. |
| Outcome: | The proposed model exceeds o1-preview on competition math questions by up to 27% (MATH and AIME24). |
When Life Gives You Samples: The Benefits of Scaling up Inference Compute for Multilingual LLMs (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in large language models have shifted focus toward scaling inference-time compute. |
| Approach: | They propose to scale inference-time compute in a multilingual, multi-task setting . they propose to use m-ArenaHard-v2.0 prompts to sample multiple outputs in parallel . |
| Outcome: | The proposed solutions achieve an average +6.8 jump in win-rates for 8B models on m-ArenaHard-v2.0 prompts in non-English languages against proprietary models like Gemini. |
The Best of Both Worlds: Combining Parallel and Sequential Inference Scaling via Aggregation Fine-Tuning (2026.findings-acl)
Copied to clipboard
| Challenge: | Empirical results show that AFT-trained models achieve substantial gains with test-time scaling. |
| Approach: | They introduce a supervised fine-tuning paradigm where models synthesize multiple draft responses into a single, refined answer. |
| Outcome: | Empirical results show that AFT-trained models outperform baseline models while eliminating external guidance. |
Table-R1: Inference-Time Scaling for Table Reasoning Tasks (2025.emnlp-main)
Copied to clipboard
| Challenge: | In this study, we explore inference-time scaling on table reasoning tasks. |
| Approach: | They propose a large-scale dataset of reasoning traces and a reinforcement learning with verifiable rewards approach to enable inference-time scaling on table reasoning tasks. |
| Outcome: | The proposed model matches or exceeds GPT-4.1 and DeepSeek-R1 models on diverse table reasoning tasks. |
Just Adjust One Prompt: Enhancing In-Context Dialogue Scoring via Constructing the Optimal Subgraph of Demonstrations and Prompts (2023.emnlp-main)
Copied to clipboard
| Challenge: | Using large language models as chatbots can cause hallucinations and lack of empathy, authors report . a dimension-agnostic scoring method is proposed to improve the performance of chatbot performance . |
| Approach: | They propose a dimension-agnostic scoring method that leverages in-context learning . they propose to automatically generate prompts and then request the LLM multiple times . |
| Outcome: | The proposed method outperforms baselines on five datasets. |
Dialogue Response Ranking Training with Large-Scale Human Feedback Data (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing open-domain dialog models can minimize the perplexity of target human responses . however, some human responses are more engaging than others, spawning more followup interactions . |
| Approach: | They train open-domain dialog models to minimize perplexity of target human responses . they use social media feedback data to train models to predict engaging dialog turns . |
| Outcome: | The proposed model outperforms existing models on 133M human feedback pairs . it also outperformed the conventional dialog perplexity baseline model . |
Inference-Time Scaling of Verification: Self-Evolving Deep Research Agents via Test-Time Rubric-Guided Verification (2026.findings-acl)
Copied to clipboard
Yuxuan Wan, Tianqing Fang, Zaitang LI, Yintong Huo, Wenxuan Wang, Haitao Mi, Dong Yu, Michael R. Lyu
| Challenge: | Recent advances in Deep Research Agents (DRAs) are transforming automated knowledge discovery and problem-solving. |
| Approach: | They propose an inference-time scaling of verification wherein an agent self-improves at test time by evaluating its generated answers. |
| Outcome: | The proposed model outperforms vanilla agent-as-judge and LLM judge baselines by 12%–48% in meta-evaluation F1 score. |
WebSTAR: Scalable Data Synthesis for Computer Use Agents with Step-Level Filtering (2026.acl-long)
Copied to clipboard
| Challenge: | Existing datasets rely on human demonstrations, limiting scalability. |
| Approach: | They propose a scalable data synthesis pipeline that transforms noisy rollouts into reliable supervision without human annotation. |
| Outcome: | The proposed pipeline transforms noisy rollouts into reliable supervision without human annotation. |
Elaboration-Generating Commonsense Question Answering at Scale (2023.acl-long)
Copied to clipboard
| Challenge: | elaborations are generated using language models that generate background knowledge that helps improve performance . human evaluations show that the quality of the generated ellaborations is high . |
| Approach: | They propose to finetune smaller language models to generate useful intermediate context . they compare a language model with an answer predictor and generate elaborations . human evaluations show that the quality of the generated ellaborations is high . |
| Outcome: | The proposed framework outperforms other models on commonsense questions on four commons sense benchmarks. |
OpenRLHF: A Ray-based Easy-to-use, Scalable and High-performance RLHF Framework (2025.emnlp-demos)
Copied to clipboard
Jian Hu, Xibin Wu, Wei Shen, Jason Klein Liu, Weixun Wang, Songlin Jiang, Haoran Wang, Hao Chen, Bin Chen, Wenkai Fang, null Xianyu, Yu Cao, Haotian Xu, Yiming Liu
| Challenge: | Existing RLHF frameworks face inference bottlenecks and complexity barriers restricting their accessibility for newcomers. |
| Approach: | They propose an open-source RLHF framework that can be used to train large language models. |
| Outcome: | The proposed framework achieves superior training efficiency with speedups ranging from 1.22 to 1.68 across different model sizes compared to state-of-the-art frameworks, while requiring significantly fewer lines of code for implementation. |