Sonnet 5.5 enters Agent Arena with a task cost tradeoff
New Agent Arena results place Sonnet 5.5 near the top while showing why effort settings and measured task spending matter alongside token prices.
Research
AI research explains how machine learning systems improve and where their weaknesses remain. ByteForward follows research papers, model evaluations and scientific applications with attention to methods, evidence and practical consequences. Topics include reasoning, training approaches, multimodal learning and the use of AI in scientific discovery. A promising result needs context about the dataset, comparison systems and conditions under which it was measured. We look at those limits so readers can distinguish a research finding from a capability ready for everyday use. Our AI safety and security coverage examines how researchers test reliability, misuse risks and defenses. For research that moves into products, follow the developments in AI models. Read the articles below for clear explanations of new AI research and the questions that determine whether its results will hold up beyond the paper.
Ai2 releases Olmo core 3 for large MoE training
Ai2 releases an open training stack with larger expert pools, detailed performance tests and migration changes for teams building sparse models.
Ai2 releases AstaBrief 8B for faster cited research reports
Ai2 opens its scientific report model and training data, with faster generation, older evaluation baselines and different licenses for weights and data.
GrowByData study finds fewer Reddit citations in ChatGPT
New research reports a citation decline in retail and travel prompts without establishing why.
AutoSynthData turns agent failures into training tasks
ServiceNow reports improved enterprise task performance using generated training examples.
Goodfire tests more selective biosecurity monitors for AI agents
New research examines a more selective approach to screening biological AI work.
Researchers propose repeatable evidence based AI security reviews
A new preprint describes a security assessment method that connects recorded software controls with explicit and repeatable scoring rules.
New paper examines the limits of private AI oversight
An October 1 preprint studies when an AI result can be checked without exposing the confidential information behind it.
Study finds AI agent communication links can weaken safeguards
A new preprint examines how learned connections between AI agents can change safety behavior even when the underlying models stay fixed.
Eni and Generative Bionics explore industrial humanoid testing
Eni and Generative Bionics will assess humanoid use cases, robot materials and computing support under a new research and industrial agreement.
Agility and FORT plan a wider safety system for Digit 5
Agility and FORT plan to connect Digit 5 with external safety systems through an expanded partnership that still requires definitive agreements.
Google study finds AI discovered flaws skew toward code execution
Google finds a different mix of vulnerabilities in AI attributed research, with important limits on what the comparison proves.
Google leads CDC flu forecast evaluation for 2026 season
The CDC ranks Google’s flu forecasting system first among 39 eligible models. The result concerns hospital admission forecasts across a defined season.
Google Suncatcher satellite reaches orbit for TPU tests
Google confirms contact with its Suncatcher prototype after launch. The mission will test how AI hardware handles radiation and cooling in orbit.
BootLoops brings reusable scientific tools to AI research
Matthew Schwartz releases a toolkit built around checkable calculations and human direction
Anthropic study finds robot costs limit physical automation
New research compares demonstrated robot capability with deployment economics
Ataraxos study expands hidden information AI beyond Stratego
The expanded Ataraxos study tests its approach in three more games, with separate evidence for human competition and AI benchmarks.
Independent Gemini 4 Argon tests show strong scores and uneven agent skills
Independent Gemini 4 Argon tests show a strong Vals Index result and a close Astra comparison, with important pricing and game control limits.
OpenAI outlines Private Intelligence with a fall inference preview
OpenAI outlines Private Intelligence and a fall Private Inference preview. Its safety processing documentation clarifies retention duties and protection boundaries.
Codex Security Cloud brings repository scans to the background
Codex Security Cloud scans GitHub repositories and monitors new commits. Its documentation explains validation, proposed patches and public review visibility.
Anthropic opens an AI interview study with optional public answers
Anthropic is asking Claude users what they want from AI. The study runs through October 6 and lets participants choose whether to publish their interviews.
Google and Ohio State expand AI research and campus tools
Google and Ohio State plan research access, campus AI tools and a student ambassador program. The agreement builds on the university's existing AI curriculum.
Google SynthID Bio brings AI watermarking to proteins
Google DeepMind’s SynthID Bio marks AI designed proteins while preserving function in lab tests. Here is what the study establishes and what still needs work.
Cross-model replay recovers hidden CoT traces at OpenAI, Anthropic, and Google
Encrypted chain-of-thought was supposed to hide proprietary reasoning from competitors and keep sensitive intermediate steps off the public page. A new paper shows those opaque blocks were portable enough that a weaker model in the same provider family could spit…