---
title: "AlphaSignal Agents: Deep Research News"
url: "https://alphasignal.ai/topic/agents/deep-research"
description: "Latest AI and machine learning news about Deep Research (Agents) on AlphaSignal."
---

# AlphaSignal Agents: Deep Research News

Latest AI and machine learning news about Deep Research (Agents) on AlphaSignal.

## Articles

- [BAAI's AREX Beats Models 10x Its Size on Deep Web Research](https://alphasignal.ai/news/baai-s-arex-beats-models-10x-its-size-on-deep-web-research.md) (2026-10-09)
- [Exa Labs' ATLAS Exposes How AI Agents Fake Web Research Without Searching](https://alphasignal.ai/news/exa-labs-atlas-exposes-how-ai-agents-fake-web-research-without-searching.md) (2026-10-08)
- [Epoch's InnovationEval Catches Frontier AI Agents Inflating Research Results](https://alphasignal.ai/news/epoch-s-innovationeval-catches-frontier-ai-agents-inflating-research-results.md) (2026-10-07)
- [MIT's Eko Agent Taught Itself Island Skills Through 30 Hours of Unsupervised Play](https://alphasignal.ai/news/mit-s-eko-agent-taught-itself-island-skills-through-30-hours-of-unsupervised.md) (2026-10-05)
- [EurekaBench Reveals AI Agents Still Fall 27 Points Behind Humans on Real Science](https://alphasignal.ai/news/eurekabench-reveals-ai-agents-still-fall-27-points-behind-humans-on-real-science.md) (2026-09-30)
- [Synthetic Sciences Ships OpenScience, an Open-Source AI Research Workbench With 250 Skills](https://alphasignal.ai/news/synthetic-sciences-ships-openscience-an-open-source-ai-research-workbench-with.md) (2026-09-27)
- [Exa's Agent Ultra Beats OpenAI and Anthropic at Web Research for 54% Less](https://alphasignal.ai/news/exa-s-agent-ultra-beats-openai-and-anthropic-at-web-research-for-54-less.md) (2026-09-25)
- [Stanford's Terminal-Bench-Science Exposes GPT-6 Astra Leading at 63% on Real Science Tasks](https://alphasignal.ai/news/stanford-s-terminal-bench-science-exposes-gpt-6-astra-leading-at-63-on-real.md) (2026-09-24)
- [Vals AI's Claude Opus 5.5 Agents Proved a Faster Shortest-Path Algorithm](https://alphasignal.ai/news/vals-ai-s-claude-opus-5-5-agents-proved-a-faster-shortest-path-algorithm.md) (2026-09-22)
- [Dream-RSI Cuts AI Coding Agent Calls by 162x Using Replay Worlds](https://alphasignal.ai/news/dream-rsi-cuts-ai-coding-agent-calls-by-162x-using-replay-worlds.md) (2026-09-18)
- [Vals AI's Terminal-Bench 4.0 Shows Most AI Agents Fail 70% of Expert Tasks](https://alphasignal.ai/news/vals-ai-s-terminal-bench-4-0-shows-most-ai-agents-fail-70-of-expert-tasks.md) (2026-09-16)
- [Google's Dream-RSI Cuts AI Discovery Agent Calls by 162x Using Replay](https://alphasignal.ai/news/google-s-dream-rsi-cuts-ai-discovery-agent-calls-by-162x-using-replay.md) (2026-09-16)
- [AIPOCH's Open Science Gives Researchers a Local AI Agent With 4,000 Stars](https://alphasignal.ai/news/aipoch-s-open-science-gives-researchers-a-local-ai-agent-with-4-000-stars.md) (2026-09-07)
- [Meta's AIRA₃ Beats 4,000 Human Teams to Win Kaggle Gold](https://alphasignal.ai/news/meta-s-aira3-beats-4-000-human-teams-to-win-kaggle-gold.md) (2026-09-05)
- [Qwen's E-Commerce Bench Exposes How Badly AI Agents Fail at Running a Business](https://alphasignal.ai/news/qwen-s-e-commerce-bench-exposes-how-badly-ai-agents-fail-at-running-a-business.md) (2026-09-03)
- [alphaXiv Opens OpenResearch, Turning Coding Agents Into Autonomous Scientists](https://alphasignal.ai/news/alphaxiv-opens-openresearch-turning-coding-agents-into-autonomous-scientists.md) (2026-09-02)
- [EvoMap's AutoResearch Stops AI From Faking Its Own Research Results](https://alphasignal.ai/news/evomap-s-autoresearch-stops-ai-from-faking-its-own-research-results.md) (2026-09-01)
- [Harvard and MIT's Science Sandbox Catches Claude Beating Humans While Missing the Point](https://alphasignal.ai/news/harvard-and-mit-s-science-sandbox-catches-claude-beating-humans-while-missing.md) (2026-08-31)
- [Sapient Intelligence's PRAXIST Runs Parallel AI Agents to Replace Manual Research Loops](https://alphasignal.ai/news/sapient-intelligence-s-praxist-runs-parallel-ai-agents-to-replace-manual.md) (2026-08-28)
- [Alibaba's CommerceAgentBench Reveals Top AI Agents Fail 38% of Real Business Tasks](https://alphasignal.ai/news/alibaba-s-commerceagentbench-reveals-top-ai-agents-fail-38-of-real-business.md) (2026-08-24)
- [Meta's WildArtifactBench Ditches Fixed Rubrics to Judge Real-World Agents](https://alphasignal.ai/news/meta-s-wildartifactbench-ditches-fixed-rubrics-to-judge-real-world-agents.md) (2026-08-20)
- [Agent Arena Splits Into 3 Categories, Proving No Single Model Wins](https://alphasignal.ai/news/agent-arena-splits-into-3-categories-proving-no-single-model-wins.md) (2026-08-17)
- [Engram's Legal Agent Beats Claude Opus 4.8 at One-Tenth the Cost](https://alphasignal.ai/news/engram-s-legal-agent-beats-claude-opus-4-8-at-one-tenth-the-cost.md) (2026-08-17)
- [Prime Intellect's Fable 5 Closes 82% of the Human AI Research Gap](https://alphasignal.ai/news/prime-intellect-s-fable-5-closes-82-of-the-human-ai-research-gap.md) (2026-08-15)
- [AQuA Runs a Self-Improving Trading AI With a 2.50 Sharpe](https://alphasignal.ai/news/aqua-runs-a-self-improving-trading-ai-with-a-2-50-sharpe.md) (2026-08-13)
- [Open-Source Pythia Gives AI Agents a Live View of 48 Real-World Feeds](https://alphasignal.ai/news/open-source-pythia-gives-ai-agents-a-live-view-of-48-real-world-feeds.md) (2026-08-08)
- [Why 99%-Accurate Agents Fail Long Horizon Tasks](https://alphasignal.ai/news/why-99-accurate-agents-fail-long-horizon-tasks.md) (2026-08-07)
- [Tencent Ships Hyra-1.0, an Agent That Critiques and Rewrites Its Own Work](https://alphasignal.ai/news/tencent-ships-hyra-1-0-an-agent-that-critiques-and-rewrites-its-own-work.md) (2026-07-21)
- [LangChain's Open Deep Research Beats Black-Box Rivals at a Fraction of the Cost](https://alphasignal.ai/news/langchain-s-open-deep-research-beats-black-box-rivals-at-a-fraction-of-the-cost.md) (2026-07-17)
- [Google Cloud's Gemini Platform Adds Parallel Web Search to Cut Agent Hallucinations](https://alphasignal.ai/news/google-cloud-s-gemini-platform-adds-parallel-web-search-to-cut-agent.md) (2026-07-16)
- [Perplexity's Open-Source WANDR Benchmark Exposes How Badly Research Agents Fail at Scale](https://alphasignal.ai/news/perplexity-s-open-source-wandr-benchmark-exposes-how-badly-research-agents-fail.md) (2026-07-14)
- [Claude Fable 5 Tops Legal AI Benchmark but Fails 86% of Real Tasks](https://alphasignal.ai/news/claude-fable-5-tops-legal-ai-benchmark-but-fails-86-of-real-tasks.md) (2026-07-07)
- [Google Cloud's ScientistOne Writes AI Research Papers With Zero Hallucinated References](https://alphasignal.ai/news/google-cloud-s-scientistone-writes-ai-research-papers-with-zero-hallucinated.md) (2026-07-07)
- [OpenAI's GeneBench-Pro Exposes That Top AI Fails Real Biology 70% of the Time](https://alphasignal.ai/news/openai-s-genebench-pro-exposes-that-top-ai-fails-real-biology-70-of-the-time.md) (2026-06-30)
- [Sakana AI's CoffeeBench Catches Claude Haiku 4.5 Going Bankrupt Over 90 Days](https://alphasignal.ai/news/sakana-ai-s-coffeebench-catches-claude-haiku-4-5-going-bankrupt-over-90-days.md) (2026-06-26)
- [Artificial Analysis' AA-Briefcase Reveals Top AI Agents Fail 97% of Real Work Tasks](https://alphasignal.ai/news/artificial-analysis-aa-briefcase-reveals-top-ai-agents-fail-97-of-real-work.md) (2026-06-24)
- [ByteDance's DeerFlow 2.0 Hits #1 on GitHub, Giving AI Agents Their Own Computer](https://alphasignal.ai/news/bytedance-s-deerflow-2-0-hits-1-on-github-giving-ai-agents-their-own-computer.md) (2026-06-19)
- [Artificial Analysis' AA-Briefcase Benchmark Exposes What AI Models Can't Actually Do](https://alphasignal.ai/news/artificial-analysis-aa-briefcase-benchmark-exposes-what-ai-models-can-t.md) (2026-06-18)
- [Exa Agent Collapses Web Research Into One API Call at Half the Cost](https://alphasignal.ai/news/exa-agent-collapses-web-research-into-one-api-call-at-half-the-cost.md) (2026-06-16)
- [You.com Rebuilt Finance AI From Scratch by Engineering Disagreement Into It](https://alphasignal.ai/news/you-com-rebuilt-finance-ai-from-scratch-by-engineering-disagreement-into-it.md) (2026-06-15)
- [Perplexity's Deep Research Rewrites Search in Python, Hits 83.8% on BrowseComp](https://alphasignal.ai/news/perplexity-s-deep-research-rewrites-search-in-python-hits-83-8-on-browsecomp.md) (2026-06-11)
- [Perplexity Computer Cuts Knowledge Work Time by 87% in Harvard Study](https://alphasignal.ai/news/perplexity-computer-cuts-knowledge-work-time-by-87-in-harvard-study.md) (2026-06-08)
- [MIT Gives AI a Math Test to Prove It's Truly Discovering](https://alphasignal.ai/news/mit-gives-ai-a-math-test-to-prove-it-s-truly-discovering.md) (2026-06-05)
- [LM Arena Ships Agent Mode to Rank AI Models on Real Multi-Step Tasks](https://alphasignal.ai/news/lm-arena-ships-agent-mode-to-rank-ai-models-on-real-multi-step-tasks.md) (2026-06-04)
- [Shanghai AI Lab's MLEvolve Beats Every MLE Agent at Half the Compute](https://alphasignal.ai/news/shanghai-ai-lab-s-mlevolve-beats-every-mle-agent-at-half-the-compute.md) (2026-06-04)
- [UC Berkeley's Agents' Last Exam Reveals AI Fails 97% of Real Professional Tasks](https://alphasignal.ai/news/uc-berkeley-s-agents-last-exam-reveals-ai-fails-97-of-real-professional-tasks.md) (2026-06-03)
- [LeanMarathon Autonomously Proves 258 Math Theorems Where Commercial AI Failed](https://alphasignal.ai/news/leanmarathon-autonomously-proves-258-math-theorems-where-commercial-ai-failed.md) (2026-06-03)
- [AutoLab Reveals Claude Opus Beats Every AI on 40-Hour Engineering Tasks](https://alphasignal.ai/news/autolab-reveals-claude-opus-beats-every-ai-on-40-hour-engineering-tasks.md) (2026-06-03)
- [IIT's Neetyabhas Trains AI to Handle Messy Pandemic Data and Non-Compliance](https://alphasignal.ai/news/iit-s-neetyabhas-trains-ai-to-handle-messy-pandemic-data-and-non-compliance.md) (2026-06-03)
- [Google DeepMind's Co-Scientist Cuts Hypothesis Discovery From Years to Days](https://alphasignal.ai/news/google-deepmind-s-co-scientist-cuts-hypothesis-discovery-from-years-to-days.md) (2026-06-02)