Papers with loss
Gatekeeper to save COGS and improve efficiency of Text Prediction (2023.emnlp-industry)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) require massive amounts of computation and storage, such an approach incurs network and high execution cost. |
| Approach: | They propose a model gatekeeper to stop LLM calls that result in incorrect predictions . they show it can save 46.6% of COGS and improve user experience by not showing incorrect predictions. |
| Outcome: | The proposed model gatekeeper saves 46.6% of COGS and improves user experience . it also improves the suggestion rate of the proposed model by 73% . |
A Strong and Robust Baseline for Text-Image Matching (P19-2)
Copied to clipboard
| Challenge: | Text-image matching is one of the most popular methods for training text-image embeddings. |
| Approach: | They propose to use a kNN-margin loss that utilizes hard negatives and is robust to noise . they advocate using Inverted Softmax and Cross-modal Local Scaling during inference . |
| Outcome: | The proposed loss function is robust to noise and pseudo negatives are tolerable . the proposed loss functions improve scores of all metrics by a large margin . |
MobileNMT: Enabling Translation in 15MB and 30ms (2023.acl-industry)
Copied to clipboard
| Challenge: | Existing work on NMT models is limited in storage, memory, computation and power consumption. |
| Approach: | They propose a mobile machine translation system that can translate in 15MB and 30ms on devices. |
| Outcome: | The proposed system can translate in 15MB and 30ms on mobile devices. |
Sparse-to-Dense: A Free Lunch for Lossless Acceleration of Video Understanding in LLMs (2025.acl-short)
Copied to clipboard
| Challenge: | Recent advances in Video Large Language Models (Video-LLMs) have achieved exceptional performance on tasks like video question answering and captioning. |
| Approach: | They propose a decoding strategy that leverages sparse top-K attention and dense full attention to accelerate Video-LLMs without loss. |
| Outcome: | The proposed approach achieves a 1.94 walltime speedup in video processing. |
TRACE: Training and Inference-Time Interpretability Analysis for Language Models (2025.emnlp-demos)
Copied to clipboard
| Challenge: | Existing tools for interpretability analysis of transformer models are post hoc, rely on scalar metrics or require nontrivial integration effort. |
| Approach: | They propose a modular toolkit for training and inference-time interpretability analysis of transformer models. |
| Outcome: | Experiments with autoregressive transformers show that TRACE reveals developmental phenomena overlooked by traditional scalar metrics such as loss or accuracy. |
Neural Latent Extractive Document Summarization (D18-1)
Copied to clipboard
| Challenge: | Existing summarization paradigms focus on extractive summarizing based on sentence level labels . |
| Approach: | They propose a latent variable extractive model where sentences are viewed as latent variables and sentences with activated variables are used to infer gold summaries. |
| Outcome: | The proposed model outperforms a strong extractive baseline trained on rule-based labels and performs competitively with several recent models. |
Improving the Robustness of Summarization Models by Detecting and Removing Input Noise (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Abstractive summarization models are typically evaluated using test data that is identically distributed as training data. |
| Approach: | They propose a method to detect and remove input noise from documents to be summarized without extra training or auxiliary models. |
| Outcome: | The proposed method recovers a large fraction of the loss in performance, sometimes as large as 11 ROUGE-1 points, without extra training, auxiliary models, or prior knowledge of the type of noise. |
Semi-Supervised Learning for Video Captioning (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Existing video captioning algorithms are heavily dependent on supervised training data. |
| Approach: | They propose to train the video captioning model on labeled and unlabeled data jointly in a semi-supervised learning manner. |
| Outcome: | The proposed model outperforms state-of-the-art semi-supervised learning approaches on VATEX, MSR-VTT and MSVD datasets. |
HuCurl: Human-induced Curriculum Discovery (2023.acl-long)
Copied to clipboard
| Challenge: | Existing curriculum learning frameworks can be used to discover effective curricula for NLP tasks based on prior knowledge about sample difficulty. |
| Approach: | They propose a framework for curriculum learning based on prior knowledge about sample difficulty. |
| Outcome: | The proposed framework outperforms existing curriculum learning approaches on several NLP tasks and can prune and weight samples for better learning. |
Using a Penalty-based Loss Re-estimation Method to Improve Implicit Discourse Relation Classification (2020.coling-main)
Copied to clipboard
| Challenge: | inessential words are unintentionally misjudged as attention-worthy words and assigned heavier attention weights than should be. |
| Approach: | They propose a penalty-based method to regulate the attention learning process by integrating penalty coefficients into the computation of loss by means of overstability of attention weight distributions. |
| Outcome: | The proposed method improves on the Penn Discourse TreeBank corpus and is competitive compared to the state-of-the-art methods. |
ECHO-LLaMA: Efficient Caching for High-Performance LLaMA Training (2025.emnlp-industry)
Copied to clipboard
Maryam Dialameh, Rezaul Karim, Hossein Rajabzadeh, Omar Mohamed Awad, Boxing Chen, Hyock Ju Kwon, Walid Ahmed, Yang Liu
| Challenge: | ECHO-LLaMA transforms LLa MA models into shared KV caching across certain layers, significantly reducing KV computational complexity while maintaining or improving language performance. |
| Approach: | They propose an efficient LLaMA architecture that transforms LLama models into shared KV caching across certain layers, reducing computational complexity while maintaining or improving language performance. |
| Outcome: | ECHO-LLaMA achieves up to 77% higher token-per-second throughput during training, up to 16% higher Model FLOPs Utilization (MFU) and up to 14% lower loss when trained on an equal number of tokens. |
On Training Data Influence of GPT Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | generative language models have redefined performance standards across tasks . current research on the influence of training data on autoregressivity remains underexplored . |
| Approach: | They propose a parameterized simulation to assess the impact of training examples on the training dynamics of GPT models. |
| Outcome: | The proposed approach compares existing methods with existing methods across training scenarios in generative language models, spanning tasks across 14 million to 2.8 billion parameters. |
Evaluating morphological typology in zero-shot cross-lingual transfer (2021.acl-long)
Copied to clipboard
| Challenge: | morphological typology has been used to improve cross-lingual transfer . however, some language families and typologies consistently perform worse . |
| Approach: | They examine effects of morphological typology on zero-shot cross-lingual transfer . they perform part-of-speech tagging and sentiment analysis on 19 languages . |
| Outcome: | The proposed model improves on fusional and introflexive languages, but some language families and typologies perform worse. |
From Stance to Concern: Adaptation of Propositional Analysis to New Tasks and Domains (2022.findings-acl)
Copied to clipboard
| Challenge: | Existing paradigms for propositional analysis use stances and concerns to generate explanatory representations. |
| Approach: | They propose a generalized paradigm for adaptation of propositional analysis to new tasks and domains by using an analogy between stances and concerns. |
| Outcome: | The proposed model yields 231% improvement in recall over baseline, with only 10% loss in precision. |
Fine- and Coarse-Granularity Hybrid Self-Attention for Efficient BERT (2022.acl-long)
Copied to clipboard
| Challenge: | Transformer-based pre-trained models achieve state-of-the-art results, but they can be prohibitively costly. |
| Approach: | They propose a fine- and coarse-granularity hybrid self-attention that shortens the computational sequence length in self- attention by progressively shortening the computational time. |
| Outcome: | The proposed model reduces computation cost by shortening the computational sequence length in self-attention. |
Towards Operationalizing Right to Data Protection (2025.naacl-long)
Copied to clipboard
| Challenge: | Recent work introduces the concept of generating unlearnable datasets (by adding imperceptible spurious correlations to the clean data) this approach is limited by several practical constraints like requiring knowledge of the target model. |
| Approach: | They propose a framework that injects imperceptible spurious correlations into natural language datasets, rendering them unlearnable without affecting semantic content. |
| Outcome: | The proposed framework can restrict newer models like GPT-4o and Llama from learning on generated data, resulting in a drop in test accuracy compared to their zero-shot performance. |
Learning to Rank for Plausible Plausibility (P19-1)
Copied to clipboard
| Challenge: | Existing approaches to encoding sentences using contextualized encoders are inconsistent . |
| Approach: | They propose to use a cross entropy log-loss objective to improve plausibility . they propose a margin-based loss leads to a more plausible model of plausability . |
| Outcome: | The proposed loss is intuitively wrong when applied to plausibility tasks . the proposed loss leads to a more plausible model of plausability . |
Between Circuits and Chomsky: Pre-pretraining on Formal Languages Imparts Linguistic Biases (2025.acl-long)
Copied to clipboard
| Challenge: | Pretraining language models on formal language can improve their acquisition of natural language, but it is computationally limited. |
| Approach: | They hypothesize that effective transfer occurs when the formal language captures hierarchical dependencies present in natural language and remains within the computational limitations of the model architecture. |
| Outcome: | The proposed model can achieve lower loss on natural language and better generalization with a 33% smaller token budget than training on a matched amount of natural language. |
Powerformer: Efficient and High-Accuracy Privacy-Preserving Language Model with Homomorphic Encryption (2025.acl-long)
Copied to clipboard
| Challenge: | a new privacy-preserving language model, Powerformer, is designed to reduce computation overhead while maintaining model performance. |
| Approach: | They propose an efficient homomorphic encryption-based privacy-preserving language model . it incorporates three key techniques to optimize encrypted computations . |
| Outcome: | The proposed model achieves 45% reduction in computation time compared to state-of-the-art models . authors say the model preserves data privacy and AI capabilities in MLaaS environments . |
Demystifying Synthetic Data in LLM Pre-training: A Systematic Study of Scaling Laws, Benefits, and Pitfalls (2025.emnlp-main)
Copied to clipboard
Feiyang Kang, Newsha Ardalani, Michael Kuchnik, Youssef Emad, Mostafa Elhoushi, Shubhabrata Sengupta, Shang-Wen Li, Ramya Raghavendra, Ruoxi Jia, Carole-Jean Wu
| Challenge: | a large-scale empirical study compares natural web data, diverse synthetic types, and mixtures of natural and synthetic data. |
| Approach: | They conduct a large-scale empirical study on large-volume LLMs using a unified protocol and scaling laws. |
| Outcome: | The proposed method is faster than pre-training on natural web data, the authors show . their results are consistent with previous studies on rephrased text and textbooks . |
Parameter-Efficient Finetuning for Robust Continual Multilingual Learning (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to Continual Multilingual Learning (CML) are based on updating models using new data in stages. |
| Approach: | They propose a parameter-efficient finetuning strategy to increase the number of languages on which the model improves after an update while reducing the magnitude of loss for the remaining languages. |
| Outcome: | The proposed model improves on the languages included in the latest update while reducing the loss of performance on the remaining languages. |
Continual Quantization-Aware Pre-Training: When to transition from 16-bit to 1.58-bit pre-training for BitNet language models? (2025.findings-acl)
Copied to clipboard
| Challenge: | Quantization-aware training of large language models reduces the precision of model parameters and reduces memory usage and energy consumption at inference time. |
| Approach: | They propose a method where models are first trained with 16-bit precision and then transition to 1.58-bit quantization-aware training. |
| Outcome: | The proposed training strategy reduces memory and energy consumption while maintaining model accuracy while reducing memory and inference time. |
Semantic Aware Linear Transfer by Recycling Pre-trained Language Models for Cross-lingual Transfer (2025.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are increasingly incorporating multilingual capabilities, fueling the demand to transfer them into target language-specific models. |
| Approach: | They propose a novel cross-lingual transfer technique that recycles embeddings from target language Pre-trained Language Models to transmit deep representational strengths to LLMs. |
| Outcome: | The proposed technique outperforms existing methods in cross-lingual understanding setups and achieves faster convergence and lower loss during language adaptation. |
Identifying Pre-training Data in LLMs: A Neuron Activation-Based Detection Framework (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for detecting pre-training data in large language models rely on superficial features like prediction confidence and loss, resulting in mediocre performance. |
| Approach: | They propose a new algorithm to analyze neuron activation patterns between training and non-training data in large language models to improve their performance. |
| Outcome: | The proposed algorithm outperforms existing methods across three benchmarks and multiple LLMs. |
CAC-CoT: Connector-Aware Compact Chain-of-Thought for Efficient Reasoning Data Synthesis Across Dual-System Cognitive Tasks (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Long chain-of-thought (CoT) prompting often slows or even degrades performance on fast, intuitive "System-1" tasks. |
| Approach: | They introduce a method that deliberately restricts reasoning to a small, fixed set of connector phrases, steering the model toward concise and well-structured explanations. |
| Outcome: | The method achieves 85% on GSM8K and 40% on GPQA while also surpassing the baseline by over 20%. |
Capability Salience Vector: Fine-grained Alignment of Loss and Capabilities for Downstream Task Scaling Law (2025.acl-long)
Copied to clipboard
Qiming Ge, Shuhao Xing, Songyang Gao, Yunhua Zhou, Yicheng Zou, Songyang Zhang, Zhi Chen, Hang Yan, Qi Zhang, Qipeng Guo, Kai Chen
| Challenge: | Large language models have demonstrated impressive performance across a wide range of tasks, but this achievement comes with the trade-off of significant computational demands. |
| Approach: | They propose a scaling law that decomposes the overall validation loss and assigns different importance weights to tokens to assess a specific meta-capability. |
| Outcome: | The proposed model can predict the loss trending of models across different levels of computation without a gap between validation loss and model's downstream capabilities. |
Initialization of Large Language Models via Reparameterization to Mitigate Loss Spikes (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to train large language models that require a non-uniform model norm are not effective. |
| Approach: | They propose a technique that allows for uniformity of the norm of the model parameters . they propose 'weight scaling as reparameterization' to adjust the norm to the parameter . |
| Outcome: | The proposed technique outperforms existing methods and stabilizes training with the transformer decoders. |
Enhancing RAG Efficiency with Adaptive Context Compression (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods apply fixed compression rates, over-compressing simple queries or under-compressed complex ones. |
| Approach: | a new framework uses a hierarchical compressor and a context selector to optimize inference efficiency . a framework that dynamically adjusts compression rates based on input complexity optimizes inference without loss of accuracy. |
| Outcome: | Adaptive Context Compression for RAG outperforms fixed-rate methods on Wikipedia and five QA datasets . |
Training Dynamics Underlying Language Model Scaling Laws: Loss Deceleration and Zero-Sum Learning (2025.acl-long)
Copied to clipboard
| Challenge: | Increasing language model size improves cross-entropy loss with power-law behaviour, but scaling laws do not explain how scaling improves loss. |
| Approach: | They find that language models undergo loss deceleration early in training . they attribute loss deceleration to a type of degenerate training dynamics we call zero-sum learning . |
| Outcome: | The proposed scaling improves loss on language models, but degrades loss in other subsets, resulting in bottlenecks. |
Detecting LLM Hallucination Through Layer-wise Information Deficiency: Analysis of Ambiguous Prompts and Unanswerable Questions (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) often generate confident yet inaccurate responses, introducing significant risks for deployment in safety-critical domains. |
| Approach: | They propose a method to detect model hallucination by systematic analysis of information flow across model layers. |
| Outcome: | The proposed approach improves model reliability by immediately integrating with universal LLMs without additional training or architectural modifications. |