SevenTnewS

Large Language Models

7 published articles

Benchmarks & Tests4 min read

Financial reasoning benchmark

LLMs know accounting formulas. FinIndices shows they can't apply them

LLMs can recite accounting formulas, but FinIndices, a benchmark built on real Chinese financial statements, shows they struggle to apply them. Gemini-3.1-Pro's accuracy fell from 70.70% to 38.22% when formula hints were removed, and generating tables actively degraded models' reasoning.

2026-08-20

AI Agents4 min read

Agent Memory

TEPA: for AI agents, a stale memory is worse than none

Agent memory has a falsifiability problem, a new arXiv preprint argues: stale facts stay retrievable and pollute the prompt. Its TEPA mechanism revokes superseded memories, and in drift tests naive memory scored below no memory (0.210 vs 0.309) while TEPA reached 0.950.

2026-08-19

Qwen / Alibaba4 min read

Qwen3.8-Max open weights land next week

Alibaba's most powerful model ever is going open source

Qwen3.8-Max, Alibaba's first open-weight Max-class model at 2.4 trillion parameters, hits Hugging Face and ModelScope next week. The launch reframes the open-source question: what happens when the frontier's biggest weights are free to download and test?

2026-08-06

LLMs & ModelsFeatured3 min read

Open Source AI

Kimi K3 is the biggest open model ever. It's still not the best.

Kimi K3 is the largest open model at 2.8T parameters, but it fails to beat the best proprietary models on overall benchmarks. While it excels on specific coding and agentic tasks, the gap to frontier leaders like Claude Fable 5 and GPT 5.6 Sol exposes the limits of scaling without architectural and data breakthroughs.

2026-07-27

AIFeatured4 min read

Philosophy of AI

No, AI is not a rival mind. It is an extension of ours

Drawing on Husserl's phenomenology, researchers argue that AI systems are best understood as extensions of natural intelligence, not as autonomous minds. This perspective explains hallucinations and compositional failures while shifting safety debates from rogue AI fears to responsible engineering and governance.

2026-07-09

Health3 min read

Identity & Privacy

The life the AI chooses to tell about you will not be yours

An AI-generated 17-slide summary of your life sounds like a neat productivity trick. The uncomfortable question is what we stop asking ourselves when a machine tells us who we are.

2026-07-07

NLP & ML4 min read

Synthetic Data

Synthetic data generation just ditched its weakest link: the boss

Matrix is a decentralized framework that uses serialized messages passed through distributed queues for multi-agent synthetic data generation. By eliminating the central orchestrator, it achieves 2 to 15 times higher throughput on identical hardware.

2026-06-06