Papers with BBH
Multi-Agent Reasoning Improves Compute Efficiency: Pareto-Optimal Test-Time Scaling (2026.acl-srw)
Copied to clipboard
| Challenge: | Inference methods that prioritize raw performance over cost-effective compute usage are not efficient for real-world applications. |
| Approach: | They evaluate inference scaling strategies to determine their computational efficiency tradeoffs . they find debate and mixture-of-agents outperform self-consistency by 1.3% and 2.7% points . |
| Outcome: | The proposed scaling strategies outperform self-consistency, self-refinement, multi-agent debate and mixture-of-a agents on reasoning tasks. |
Divide, Reweight, and Conquer: A Logit Arithmetic Approach for In-Context Learning (2026.eacl-long)
Copied to clipboard
| Challenge: | In-Context Learning (ICL) is a key feature for Large Language Models (LLMs) but it faces challenges when dealing with increasing numbers of examples due to performance degradation and quadratic computational costs. |
| Approach: | They propose a Logit Arithmetic Reweighting Approach that uses logit-based ensembling of multiple demonstrations to enhance ICL. |
| Outcome: | The proposed framework outperforms baseline methods in accuracy and memory efficiency. |
LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression (2024.findings-acl)
Copied to clipboard
Zhuoshi Pan, Qianhui Wu, Huiqiang Jiang, Menglin Xia, Xufang Luo, Jue Zhang, Qingwei Lin, Victor Rühle, Yuqing Yang, Chin-Yew Lin, H. Vicky Zhao, Lili Qiu, Dongmei Zhang
| Challenge: | Existing approaches to compress prompts only leverage unidirectional context, causing suboptimal results. |
| Approach: | They propose a task-agnostic prompt compression method that takes tokens from context . they use a Transformer encoder to capture all essential information needed for prompt compression . |
| Outcome: | The proposed method is 3x-6x faster than existing prompt compression methods and faster than baselines. |
Autonomous Data Selection with Zero-shot Generative Classifiers for Mathematical Texts (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods that require human annotations or training a dedicated data filter to curate high-quality mathematical texts are based on autonomous data selection. |
| Approach: | They propose a method that leverages base language models as zero-shot "generative classifiers" they use a model's logits to determine whether a given passage is mathematically informative and educational . |
| Outcome: | The proposed method significantly boosts downstream performance on math benchmarks while using far fewer tokens than previous methods. |
SyncThink: A Training-Free Strategy to Align Inference Termination with Reasoning Saturation (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) achieve strong reasoning with Chain-of-Thought prompting, but long and redundant traces substantially increase inference cost. |
| Approach: | They propose a training-free and plug-and-play decoding method that reduces CoT overhead without modifying model weights. |
| Outcome: | Experiments on GSM8K, MMLU, GPQA, and BBH show that SyncThink achieves 62.00% average Top@1 accuracy using 656 generated tokens and 28.68s latency, compared to 61.22%, 2141 tokens, and 92.01s for full CoT decoding. |
Self-Reasoning Language Models: Unfold Hidden Reasoning Chains with Few Reasoning Catalyst (2025.findings-acl)
Copied to clipboard
| Challenge: | Recent studies have demonstrated that inference-time scaling increases performance of Large Language Models (LLMs) in various reasoning tasks such as mathematics and complex question answering by increasing the length of Chain-of-Thought (CoT). |
| Approach: | They propose a model which synthesizes longer CoT data and iteratively improves performance through self-training by incorporating a few demonstration examples. |
| Outcome: | The proposed model achieves an average improvement of more than +2.5 points across five reasoning tasks: MMLU, GSM8K, ARC-C, HellaSwag, and BBH on two backbone models. |
Teaching LLMs to Plan, Not Just Solve: Plan Learning Boosts LLMs Generalization in Reasoning Tasks (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for reinforcement learning (RL) on self-generated data are limited in many domains. |
| Approach: | a new framework combines plan-based search with Step-level Advantage Preference Optimization to optimize plan learning. |
| Outcome: | The proposed framework improves in-domain performance and out-of-domain benchmarks. |
Reasoning Robustness of LLMs to Adversarial Typographical Errors (2024.emnlp-main)
Copied to clipboard
Esther Gan, Yiran Zhao, Liying Cheng, Mao Yancan, Anirudh Goyal, Kenji Kawaguchi, Min-Yen Kan, Michael Shieh
| Challenge: | Large Language Models (LLMs) have demonstrated impressive capabilities in reasoning using Chain-of-Thought (CoT) prompting. |
| Approach: | They develop an algorithm that iteratively samples typos for words that are important to the query and selects the edit that is most likely to succeed in attacking. |
| Outcome: | The proposed algorithm detects typographical errors in large and closed-source LLMs and shows that they are robust to them. |
Learning from Mistakes via Cooperative Study Assistant for Large Language Models (2023.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) have demonstrated their potential to refine their generation based on their own feedback, but the feedback from LLM itself is often inaccurate, thereby limiting its benefits. |
| Approach: | They propose a framework with an auxiliary agent to assist the main LLM in learning from mistakes through interactive cooperation. |
| Outcome: | The proposed framework can significantly boost large language models by an accuracy margin of up to 6.6 on BBH and 12.6 on BBQ. |
Harder Task Needs More Experts: Dynamic Routing in MoE Models (2024.acl-long)
Copied to clipboard
Quzhe Huang, Zhenwei An, Nan Zhuang, Mingxu Tao, Chen Zhang, Yang Jin, Kun Xu, Kun Xu, Liwei Chen, Songfang Huang, Yansong Feng
| Challenge: | Unlike existing MoE approaches that rely on fixed TopK Routing, our dynamic expert selection framework dynamically allocates experts based on the confidence level in expert selection for each input. |
| Approach: | They propose a dynamic expert selection framework that dynamically allocates experts based on the confidence level in expert selection for each input. |
| Outcome: | The proposed method achieves an average improvement of 0.7% with less than 90% activated parameters and outperforms dense models in QA and machine translation tasks. |
PMPO: Probabilistic Metric Prompt Optimization for Small and Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods evaluate candidate prompts by sampling full outputs, often coupled with self critique or human annotated preferences, which limits scalability, especially for smaller models or models that are not instruction tuned. |
| Approach: | They propose a framework that uses token level cross entropy as a direct, lightweight evaluation signal to evaluate candidate prompts. |
| Outcome: | The proposed framework outperforms prior prompt optimizers across model sizes and datasets. |
Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them (2023.findings-acl)
Copied to clipboard
Mirac Suzgun, Nathan Scales, Nathanael Schärli, Sebastian Gehrmann, Yi Tay, Hyung Won Chung, Aakanksha Chowdhery, Quoc Le, Ed Chi, Denny Zhou, Jason Wei
| Challenge: | Language models have already made good progress on this benchmark, with the best model outperforming average reported human-rater results on 65% of the BIG-Bench tasks. |
| Approach: | They propose to use chain-of-thought prompting to challenge language models on 23 challenging BIG-Bench tasks which they call BIG-Bench Hard. |
| Outcome: | The proposed language models outperform the average human-rater on 65% of the BIG-Bench tasks. |
LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models (2023.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) are increasingly lengthy and require longer prompts . this paper presents a coarse-to-fine prompt compression method to reduce cost and increase performance. |
| Approach: | They propose a coarse-to-fine prompt compression method that maintains semantic integrity under high compression ratios and a token-level iterative compression algorithm to better model the interdependence between compressed contents. |
| Outcome: | The proposed method yields state-of-the-art performance and allows for up to 20x compression with little performance loss over four datasets from different scenarios. |
DAC: A Dynamic Attention-aware Approach for Task-Agnostic Prompt Compression (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods rely on information entropy as the metric to compress lexical units, but ignore attention-critical tokens and information . recent advent of In-Context Learning (ICL), Chain-of-Thought (CoT), and Retrieval Augmented Generation (RAG) technologies has significantly invigorated the landscape of applications based on Large Language Models (LLMs). |
| Approach: | They propose a dynamic attention-aware approach to task-agnostic prompt compression . they integrate entropy and attention information to achieve fine-grained prompt compression. |
| Outcome: | Experiments show that the proposed approach improves across tasks and LLMs. |
HyperLoRA: Efficient Cross-task Generalization via Constrained Low-Rank Adapters Generation (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to adapt pre-trained language models (PLMs) to emerging tasks are costly and inefficient. |
| Approach: | They propose a meta-network that generates task-specific weights without any optimization. |
| Outcome: | The proposed approach has flexible generalization ability and superior performance over hypenetworks. |
DEM: Distribution Edited Model for Training with Mixed Data Distributions (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent fine-tuning approaches for large language models require supervised finetun on diverse datasets and follow different distributions. |
| Approach: | They propose a distribution edited model that integrates models individually trained on each data source with the base model using basic element-wise vector operations. |
| Outcome: | The proposed model outperforms baseline models on a variety of benchmarks and is cheaper than standard data mixing methods. |
Triangulating LLM Progress through Benchmarks, Games, and Cognitive Tests (2025.findings-emnlp)
Copied to clipboard
Filippo Momentè, Alessandro Suglia, Mario Giulianelli, Ambra Ferrari, Alexander Koller, Oliver Lemon, David Schlangen, Raquel Fernández, Raffaella Bernardi
| Challenge: | MMLU and BBH are three evaluation paradigms for language learning models . interactive games are superior to standard benchmarks in discriminating models based on human cognitive assessments . |
| Approach: | They examine three evaluation paradigms: standard benchmarks, interactive games and cognitive tests . they examine whether interactive games are more effective at discriminating LLMs . |
| Outcome: | The results show that interactive games are superior to standard benchmarks in discriminating models. |
BIG-Bench Extra Hard (2025.acl-long)
Copied to clipboard
Mehran Kazemi, Bahare Fatemi, Hritik Bansal, John Palowitch, Chrysovalantis Anastasiou, Sanket Vaibhav Mehta, Lalit K Jain, Virginia Aglietti, Disha Jindal, Peter Chen, Nishanth Dikkala, Gladys Tyen, Xin Liu, Uri Shalit, Silvia Chiappa, Kate Olszewska, Yi Tay, Vinh Q. Tran, Quoc V Le, Orhan Firat
| Challenge: | Current benchmarks for large language model reasoning focus on math and coding abilities, leaving a gap in evaluating broader reasoning proficiencies. |
| Approach: | They propose a benchmark to evaluate general reasoning in large language models . they use BIG-Bench and its harder version BIG-Benefit Hard to assess general reasoning . |
| Outcome: | The new benchmark pushes the boundaries of LLM reasoning evaluation. |
Probing and Boosting Large Language Models Capabilities via Attention Heads (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to identifying capabilities rely on external signals with limited structural grounding . emergence of specific capabilities remains poorly understood . |
| Approach: | They propose a lightweight approach that links LLM capabilities to internal components by identifying correspondences at the level of attention heads. |
| Outcome: | The proposed approach improves accuracy on MMLU and BBH by 1 to 1.5 points over gradient-based method and 5 to 6 points over other intermediate-state baselines. |