Natural Language Processing RSS Feeds

Directory3 natural language processing feeds · 34 articlesShow all languages →

Latest

Natural Language Processing articles

LLM-INSTRUCT: Constraint-Aware Retrieval and Selective Debate for Paragraph-Level Argument Mining

We present LLM-INSTRUCT, the winning system for the UZH Shared Task at ArgMining 2026 on paragraph-level argument mining in UN and UNESCO resolutions. The task requires paragraph-type classification, prediction of a subset of 141 official tags, and direct

Insighttechnologyscienceartificial intelligence

arxiv.org·Tech by Flipboard·25d ago·2 min read

SALR: A Sparsity-Aware Low-Rank Fine-Tuning Method That Reduces LLM Model Size by 2x While Maintaining Performance

Adapting large pre-trained language models to downstream tasks often entails fine-tuning millions of parameters or deploying costly dense weight updates, which hinders their use in resource-constrained environments. Low-rank Adaptation (LoRA) reduces trai

Insighttechnologyscience

Shared by @ai-firehose.column.social ↗

arxiv.org28d ago

Building bilingual NER for cargo logistics using Amazon Bedrock knowledge distillation

Insighttechnologyartificial intelligence

Shared by @sergiocuellar.bsky.social ↗

aws.amazon.com1mo ago

colibri-embed-ptbr: A Compact 157M-Parameter Brazilian Portuguese Embedding Model That Competes with Much Larger Systems

Insighttechnologyartificial intelligence

Shared by @hackernoon.com ↗

hackernoon.com1mo ago

CRISP: Using Sparse Autoencoders for Persistent Concept Unlearning in Neural Networks

Tomer Ashuach, Dana Arad, Aaron Mueller, Martin Tutek, Yonatan Belinkov. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Insighttechnologysciencemachine learning

aclanthology.org·Twitter Design (trending)·1mo ago·1 min read

LLMs Fail the Imperfective Paradox: Study Reveals Teleological Bias in Semantic Event Understanding

Bolei Ma, Yusuke Miyao. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Insighttechnologysciencenatural language processing

aclanthology.org·Twitter AI (high signal)·1mo ago·2 min read

Weaver: Autoregressive drafting with factorized priors for efficient speculative decoding

Speculative decoding greatly increases the interactivity of autoregressive language models by trading off computation for extra tokens generated in a single forward pass. Factorized draft models are especially efficient because they predict future-token m

Insighttechnologymachine learningprogramming

arxiv.org·Twitter AI (high signal)·1mo ago·1 min read

Transformer LMs encode some relational knowledge linearly, but not all

Much of the knowledge encoded in transformer language models (LMs) may be expressed in terms of relations: relations between words and their synonyms, entities and their attributes, etc. We show that, for a subset of relations, this computation is well-ap

Insighttechnologyscience

Shared by @stanfordnlp ↗

arxiv.org1mo ago

Dispersion loss counteracts embedding condensation to improve small language model generalization

Insighttechnologyscience

chenliu-1996.github.io1mo ago

HOLA: Adding an Exact Memory Cache to Linear-Attention Language Models Improves Long-Context Recall

Linear-attention and state-space language models compress the prefix into a fixed-size recurrent state, yielding O(1) memory at the cost of a lossy exact memory: when many key--value associations compete, earlier facts are overwritten and needle recall de

Newstechnologyscience

Shared by @omarsar0 ↗

arxiv.org1mo ago

RAGP: A Graph-Based Prompt Compression Method Using Lévy Walk-Guided Pruning

Existing prompt compression methods treat text as flat token sequences, failing to capture the distributed nature of important information, which is often spread across multiple locations and connected through both local syntactic dependencies and global

Insighttechnologyscience

Shared by @news.karthihegde.dev ↗

arxiv.org1mo ago

Cartridges: Training smaller KV caches via self-study to replace in-context learning in LLMs

Large language models are often used to answer queries grounded in large text corpora (e.g. codebases, legal documents, or chat histories) by placing the entire corpus in the context window and leveraging in-context learning (ICL). Although current models

Insighttechnologyscience

Shared by @cwolferesearch ↗

arxiv.org1mo ago

DOPD: A Dual On-policy Distillation Method to Address Privilege Illusion in LLM and VLM Training

On-policy distillation (OPD) offers superior capacity transfer by supervising student-sampled trajectories with dense token-level signals. To furnish high-quality supervision sources and thereby elevate the performance frontier of distillation, an intuiti

Insighttechnologyscience

Shared by @teortaxesTex ↗

arxiv.org1mo ago

BINEVAL: A Binary Question Framework for Interpretable LLM Evaluation and Self-Improvement

Evaluating LLM outputs remains a major bottleneck in NLP: human evaluation is expensive and slow, lexical metrics correlate poorly with human judgments on open-ended generation, and holistic LLM judges often produce opaque scores that are hard to debug. W

Insighttechnologyscienceartificial intelligence

arxiv.org·Twitter AI (high signal)·1mo ago·2 min read

Doctolib's FineMed Medical-Entity Extractor: A French Medical NLP Model on Hugging Face

technologyhealth

Shared by @huggingface ↗

huggingface.co1mo ago

Three training-time interventions improve diffusion-based speculative decoding by 21-76%

Large language models (LLMs) achieve remarkable performance across a wide range of tasks, but their autoregressive decoding process incurs substantial inference costs due to inherently sequential token generation. Speculative decoding addresses this bottl

Insighttechnologyscience

Shared by @ai-firehose.column.social ↗

arxiv.org1mo ago

Speech and Language Processing (3rd ed. draft) - January 2026 Release

Speech and Language Processing

technologyeducation

Shared by @stanfordnlp ↗

web.stanford.edu1mo ago

Study reveals why in-context learning fails on complex specification-heavy tasks and how fine-tuning can help

In-context learning (ICL) has become the default method for using large language models (LLMs), making the exploration of its limitations and understanding the underlying causes crucial. In this paper, we find that ICL falls short of handling specificatio

Insighttechnologyscience

Shared by @ai-linkstream.bsky.social ↗

arxiv.org1mo ago

ACL Anthology2

cobusgreyling.medium.com1

LLM-INSTRUCT: Constraint-Aware Retrieval and Selective Debate for Paragraph-Level Argument Mining

SALR: A Sparsity-Aware Low-Rank Fine-Tuning Method That Reduces LLM Model Size by 2x While Maintaining Performance

Building bilingual NER for cargo logistics using Amazon Bedrock knowledge distillation

colibri-embed-ptbr: A Compact 157M-Parameter Brazilian Portuguese Embedding Model That Competes with Much Larger Systems

CRISP: Using Sparse Autoencoders for Persistent Concept Unlearning in Neural Networks

LLMs Fail the Imperfective Paradox: Study Reveals Teleological Bias in Semantic Event Understanding

Weaver: Autoregressive drafting with factorized priors for efficient speculative decoding

Transformer LMs encode some relational knowledge linearly, but not all

Dispersion loss counteracts embedding condensation to improve small language model generalization

HOLA: Adding an Exact Memory Cache to Linear-Attention Language Models Improves Long-Context Recall

RAGP: A Graph-Based Prompt Compression Method Using Lévy Walk-Guided Pruning

Cartridges: Training smaller KV caches via self-study to replace in-context learning in LLMs

DOPD: A Dual On-policy Distillation Method to Address Privilege Illusion in LLM and VLM Training

BINEVAL: A Binary Question Framework for Interpretable LLM Evaluation and Self-Improvement

Doctolib's FineMed Medical-Entity Extractor: A French Medical NLP Model on Hugging Face

Three training-time interventions improve diffusion-based speculative decoding by 21-76%

Speech and Language Processing (3rd ed. draft) - January 2026 Release

Study reveals why in-context learning fails on complex specification-heavy tasks and how fine-tuning can help