[research]

Research

127 stories · page 5

[research]Jul 16, 2026By ByteBulletin Editor

New arXiv Study Finds LLMs Have a 'Premise Dependency' Testing Blind Spot

A new paper reveals that large language models struggle when sub-tasks depend on the truth of intermediate premises, pointing to a fundamental reasoning gap.

[research]Jul 14, 2026By ByteBulletin Editor

KV Cache Compression Study Shows Potential for Dramatically Faster LLM Inference

New research on compressing the key-value cache in transformer models could reduce memory usage and latency, enabling longer context windows and cheaper deployment.

[research]Jul 11, 2026By ByteBulletin Editor

Oyster II: A New Framework for Safety Alignment in Large Models

Researchers propose Oyster II, a novel approach to align large AI models with human values, emphasizing robustness against adversarial attacks.

[research]Jul 11, 2026By ByteBulletin Editor

Recolora: Continual Fine-Tuning for LLMs Without Catastrophic Forgetting

A new method enables LLMs to adapt to new tasks or domains while retaining previously learned capabilities, addressing a key challenge in model deployment.

[research]Jul 10, 2026By ByteBulletin Editor

Twelve models, four apps, five attempts each: the TryAI build-off is back

GPT-5.6 tiers, Claude Fable, and an open-source surprise go head-to-head on raycaster, Rubik's Cube, calculator, and SVG tasks.

[research]Jul 10, 2026By ByteBulletin Editor

Prismata: A New Framework for Understanding and Mitigating Prompt Injection in LLMs

Researchers introduce a formal taxonomy and defense framework for prompt injection attacks, bridging theory and practice.

[research]Jul 10, 2026By ByteBulletin Editor

SwarmResearch: How LLM Coding Agents Collaborate on GitHub Issues

A new paper introduces a multi-agent framework where specialized LLM agents work together to autonomously resolve open-source software issues.