Papers with computational power
Theoretical Limitations of Self-Attention in Neural Sequence Models (2020.tacl-1)
Copied to clipboard
| Challenge: | Existing work suggests that the computational capabilities of self-attention to model hierarchical structures are limited. |
| Approach: | They investigate the computational power of self-attention to model formal languages . they show strong theoretical limitations of self attention to model periodic finite-state languages unless the number of layers or heads increases with input length. |
| Outcome: | The proposed models can model periodic finite-state languages, nor hierarchical structure unless the number of layers or heads increases with input length. |
The Parallelism Tradeoff: Limitations of Log-Precision Transformers (2023.tacl-1)
Copied to clipboard
| Challenge: | Existing work on transformers established their Turing completeness, albeit with assumptions like infinite precision and arbitrarily powerful feedforward subnets. |
| Approach: | They propose to simulate transformers whose arithmetic precision is logarithmic in the number of input tokens by constant-depth logspace-uniform threshold circuits. |
| Outcome: | The proposed model emerges from the transformer architecture’s high parallelizability, suggesting a potential inherent weakness of the scaling paradigm. |
Parameter-Efficient Fine-Tuning via Circular Convolution (2025.findings-acl)
Copied to clipboard
| Challenge: | Low-Rank Adaptation (LoRA) has gained popularity for fine-tuning large foundation models, but its intrinsic low-rank characteristic may limit its performance. |
| Approach: | They propose a low-rank adaptive method that uses low-ranked matrices to represent weight changes. |
| Outcome: | The proposed method reduces trainable parameters and mitigates heavy memory consumption associated with full delta matrices by sequentially multiplying mathbf A and mathbb B with the activation. |
PsFuture: A Pseudo-Future-based Zero-Shot Adaptive Policy for Simultaneous Machine Translation (2024.emnlp-main)
Copied to clipboard
| Challenge: | Simultaneous machine translation (SiMT) requires target tokens to be generated in real-time as streaming source tokens are consumed. |
| Approach: | They propose a zero-shot adaptive read/write policy for siMT that generates target tokens concurrently as streaming source tokens are consumed. |
| Outcome: | The proposed policy achieves performance on par with strong baselines and the P2F method can further enhance performance. |
On the Practical Computational Power of Finite Precision RNNs for Language Recognition (P18-2)
Copied to clipboard
| Challenge: | Recurrent Neural Networks (RNNs) are famously known to be Turing complete, but this relies on infinite precision in the states and unbounded computation time. |
| Approach: | They propose to use LSTM and Elman-RNN with ReLU activation to study RNNs . they show that LS and ReLU-RNns can easily implement counting behavior . |
| Outcome: | The LSTM and the Elman-RNN with ReLU activation are stronger than the RNN with squashing activation and the GRU. |
Data-Efficient French Language Modeling with CamemBERTa (2023.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in NLP have significantly improved the performance of language models on a variety of tasks. |
| Approach: | They introduce a French DeBERTa model that builds upon the DeBERTAV3 architecture and training objective and evaluate its performance on a variety of French downstream tasks and datasets. |
| Outcome: | The proposed model outperforms BERT-based models on most tasks given the same amount of training tokens and trained on 30% of its input tokens. |
Contribution of Move Structure to Automatic Genre Identification: An Annotated Corpus of French Tourism Websites (2024.lrec-main)
Copied to clipboard
| Challenge: | a concept of move structure has been overlooked in genre analysis, but it is not widely used in natural language processing. |
| Approach: | They propose to incorporate move structure into a neural architecture for automatic genre identification. |
| Outcome: | The proposed approach can increase performance and reduce computational power. |
On the Representational Capacity of Recurrent Neural Language Models (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies have focused on LMs as formal languages, but they do not consider language membership. |
| Approach: | They extend the Turing completeness result to the probabilistic case . they show that a rationally weighted RLM can simulate any deterministic Turing machine . |
| Outcome: | The proposed model can simulate any deterministic Turing machine with rationally weighted transitions . the proposed model is based on recurrent neural networks with a rational weighting over strings . |
Transfer Fine-Tuning: A BERT Case Study (D19-1)
Copied to clipboard
| Challenge: | Recent advances in sentence representation learning have improved the performance of BERT models, but the computational power required is an obstacle preventing practical applications from adopting the technology. |
| Approach: | They propose to inject phrasal paraphrase relations into BERT to generate suitable representations for semantic equivalence assessment instead of increasing model size. |
| Outcome: | The proposed model improves a smaller model while maintaining the model size. |
VoxCommunis: A Corpus for Cross-linguistic Phonetic Analysis (2022.lrec-1)
Copied to clipboard
| Challenge: | Until recently, the movement towards large-scale cross-linguistic phonetic research has been limited. |
| Approach: | They propose to use the VoxCommunis Corpus to facilitate cross-linguistic phonetic research . corpus contains acoustic models, pronunciation lexicons, word- and phone-level alignments . |
| Outcome: | The VoxCommunis Corpus contains acoustic models, pronunciation lexicons, word- and phone-level alignments . the corpus is free to download and use under a CC0 license . |
Practical Computational Power of Linear Transformers and Their Recurrent and Self-Referential Extensions (2023.emnlp-main)
Copied to clipboard
| Challenge: | Recent studies of the computational power of recurrent neural networks reveal a hierarchy of RNN architectures, given finite-precision assumptions. |
| Approach: | They propose to use auto-regressive Transformers with linearised attention to build RNNs . they show that many well-known results for the standard Transformer directly transfer to LTs - a new approach is proposed . |
| Outcome: | The proposed extensions overcome limitations of the LT and self-referential weight matrices. |
An Automated Framework for Fast Cognate Detection and Bayesian Phylogenetic Inference in Computational Historical Linguistics (P19-1)
Copied to clipboard
| Challenge: | Existing methods for phylogenetic reconstruction of large datasets require time and computational power. |
| Approach: | They propose a workflow for phylogenetic reconstruction on large datasets using two methods . they use a method for fast detection of cognates and a Bayesian method for inference . their results show that the methods take less than a few minutes to process language families . |
| Outcome: | The proposed methods are fast and easy to use and close to gold standard cognate judgments and expert language family trees. |
Federated Foundation Models: Privacy-Preserving and Collaborative Learning for Large Models (2024.lrec-main)
Copied to clipboard
| Challenge: | Foundation Models (FMs) have demonstrated success in a wide range of applications, but their optimization often requires access to sensitive data. |
| Approach: | They propose a framework that combines FMs and Federated Learning to enable privacy-preserving and collaborative learning across multiple end-users. |
| Outcome: | The proposed framework combines benefits of FMs and Federated Learning (FL) it enables privacy-preserving and collaborative learning across multiple end-users. |
On the Representational Capacity of Neural Language Models with Chain-of-Thought Reasoning (2024.acl-long)
Copied to clipboard
| Challenge: | Existing theoretical treatments have shed light on some aspects of CoT, but we are still far from a concrete understanding of the concept. |
| Approach: | They propose to formalize CoT reasoning in a probabilistic setting to bridge this gap . they show that LMs can represent the same family of distributions over strings as probabilistic Turing machines. |
| Outcome: | The proposed model can represent the same family of distributions over strings as Turing machines. |
Distributed LLM Serving on Consumer-Grade GPUs by Reconciling Computation and Communication (2025.findings-emnlp)
Copied to clipboard
Lewei Jin, Kui Zhang, Yongqi Chen, null Zhuoyifan, Renjie Li, Yi Gao, Bowei Yang, Zhengong Cai, Wei Dong
| Challenge: | Large language models are reshaping internet services, and serving them is costly. |
| Approach: | They propose an efficient distributed LLM serving system that splits prefill and decode requests into smaller chunks . |
| Outcome: | The proposed system reduces TTFT, TPOT, and latency compared to the state-of-the-art system. |
PK-ICR: Persona-Knowledge Interactive Multi-Context Retrieval for Grounded Dialogue (2023.emnlp-main)
Copied to clipboard
| Challenge: | Identifying relevant persona or knowledge for conversational systems is difficult, but recent work has shown that it is more realistic to optimize for concrete persona. |
| Approach: | They propose a persona-knowledge dual context retrieval method that utilizes all dialogue contexts simultaneously. |
| Outcome: | The proposed method performs zero-shot top-1 knowledge retrieval and precise persona scoring. |
Scaling Laws and Efficient Inference for Ternary Language Models (2025.acl-long)
Copied to clipboard
Tejas Vaidhya, Ayush Kaushal, Vineet Jain, Francis Couture-Harpin, Prashant Shishodia, Majid Behbahani, Yuriy Nevmyvaka, Irina Rish
| Challenge: | Large language models (LLMs) are increasingly used across research and industry applications, yet their inference efficiency remains a challenge. |
| Approach: | They propose ternary language models that employ quantization-aware training to significantly reduce memory requirements. |
| Outcome: | The proposed ternary language models demonstrate sustained performance gains at scale. |