// espacio-latente.com / radar

El radar

Píldora diaria de lo que se mueve en IA: laboratorios, papers, blogs de referencia y releases relevantes, resumido y con link al original. Se actualiza dos veces al día. · ← volver a espacio-latente.com · archivo · RSS

2026-09-07

arXiv cs.CL

How Much Does Corpus Choice Change Dependency-Distance Estimates?

How Much Does Corpus Choice Change Dependency-Distance Estimates?

arXiv cs.CL

Memory as transformation: LETHE, a self-referential gan-inspired architecture

Memory as transformation: LETHE, a self-referential gan-inspired architecture

arXiv cs.CL

Evidence Integration in Large Language Models

Evidence Integration in Large Language Models

arXiv cs.CL

MedProb: Probing Internal Representations of Vision-Language Models for Medical Question Answering

MedProb: Probing Internal Representations of Vision-Language Models for Medical Question Answering

arXiv cs.CL

Adapting from Downturns: Prediction of Long-Term Conversational-Skill Development in Mental-Health Crisis Counselors

Adapting from Downturns: Prediction of Long-Term Conversational-Skill Development in Mental-Health Crisis Counselors

arXiv cs.CL

VERGE: Verification-Enhanced Refinement for Grounded Extraction of Early-Onset Colorectal Cancer Symptoms in Clinical Notes

VERGE: Verification-Enhanced Refinement for Grounded Extraction of Early-Onset Colorectal Cancer Symptoms in Clinical Notes

arXiv cs.CL

You Really Didn't Get That? Benchmarking Social Pragmatic Inference for Indirect and Playful Chinese Online Comments

You Really Didn't Get That? Benchmarking Social Pragmatic Inference for Indirect and Playful Chinese Online Comments

arXiv cs.CL

Evaluation of Phonetic Encoding Algorithms on Transcription Datasets

Evaluation of Phonetic Encoding Algorithms on Transcription Datasets

arXiv cs.CL

The Anatomy of an ASR Hallucination

The Anatomy of an ASR Hallucination

arXiv cs.CL

A Systematic Evaluation of Cross-Lingual Consistency Enhancement Methods in Multilingual Language Models

A Systematic Evaluation of Cross-Lingual Consistency Enhancement Methods in Multilingual Language Models

arXiv cs.CL

What Attention Recalls and Recurrence Controls in Hybrid Language Models

What Attention Recalls and Recurrence Controls in Hybrid Language Models

arXiv cs.CL

GRACE: Graph-Grounded Reflective Agent Copilot Engine for Expert-in-the-Loop Knowledge Expansion

GRACE: Graph-Grounded Reflective Agent Copilot Engine for Expert-in-the-Loop Knowledge Expansion

arXiv cs.CL

TRILOGUE: A Trilingual Spoken Dialogue Fact-Checking Benchmark with Evidence and Paired Audio

TRILOGUE: A Trilingual Spoken Dialogue Fact-Checking Benchmark with Evidence and Paired Audio

arXiv cs.CL

Shared circuits predict whether LLMs generalize across formats in arithmetic reasoning

Shared circuits predict whether LLMs generalize across formats in arithmetic reasoning

arXiv cs.CL

Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal

Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal

arXiv cs.CL

Patterns of Priming in Production: Lexical, Semantic and Structural Alignment in Language Model Generation

Patterns of Priming in Production: Lexical, Semantic and Structural Alignment in Language Model Generation

arXiv cs.CL

Cultural Misalignment in Large Language Models: Detection, Measurement, and Mitigation Through Targeted Fine-Tuning

Cultural Misalignment in Large Language Models: Detection, Measurement, and Mitigation Through Targeted Fine-Tuning

arXiv cs.CL

Towards Understanding Pause Token Fine-Tuning Dynamics: A Mode Retention Perspective

Towards Understanding Pause Token Fine-Tuning Dynamics: A Mode Retention Perspective

arXiv cs.CL

LentEx: Generalizable Latent Entity Extraction via Synthetic Data and Instruction-Tuned LLMs

LentEx: Generalizable Latent Entity Extraction via Synthetic Data and Instruction-Tuned LLMs

arXiv cs.CL

Scale-QLoRA: Code-Invariant Adapter Merging for Native 4-bit Microscaling LLMs

Scale-QLoRA: Code-Invariant Adapter Merging for Native 4-bit Microscaling LLMs

OpenAI News

An Alien Mind

An Alien Mind

Simon Willison

Research acceleration: The view inside OpenAI

Research acceleration: The view inside OpenAI

Simon Willison

The purpose of DNS is to spread scams

The purpose of DNS is to spread scams

Simon Willison

There's No Limit to How Bad Code Can Get

There's No Limit to How Bad Code Can Get

Simon Willison

Quoting Zach Kehs

Quoting Zach Kehs

Construir este día costó 45.540 tokens de entrada y 0 de salida en 25 llamadas a modelos — unos 0,0064 $.