evald.ai

MLCommons Evaluation Filter · September 17, 2026 15:00

Where the Industry Is Investing: A Look at MLPerf Inference v6.1 - MLCommons

MLPerf Inference Working Group chairs Miro Hodak and Frank Han analyze the broadest round yet — 30 submitters, 120 systems, new agentic benchmarks, and record performance leaps.

Topics: Benchmarks

Entities: Benchmarks

OpenAI Evaluation Filter · September 16, 2026 17:00

Our framework for reporting model misalignment

OpenAI shares a framework for tracking, investigating, and disclosing model misalignment, alongside six reports of unexpected or concerning model behavior.

Entities: OpenAI

MLCommons Evaluation Filter · September 16, 2026 14:45

MLCommons Sets Participation Record with New MLPerf Inference v6.1 Benchmark Results - MLCommons

Latest version of the benchmark introduces two new tests for emerging AI deployment patterns, including Agentic Inference

Topics: Benchmarks

Entities: Benchmarks

Mitchell Bryson AI Reliability Articles · September 13, 2026 00:00

AI safety is finally testing the business plan - Mitchell Bryson

OpenAI rules out a 2026 IPO while Anthropic promises permanent outside evaluators and reportedly seeks Nvidia's backing for its flotation. Our argument: product builders should judge safety commitments by the independent scrutiny and commercial constraints...

Topics: Safety EvalsTesting Tools

Entities: AnthropicSafety EvalsTesting ToolsOpenAINVIDIA

Mitchell Bryson AI Reliability Articles · September 13, 2026 00:00

The slider changes what you sell - Mitchell Bryson

Prima Frutta uses TOMRA’s LUCAi to preview how setting changes affect cherry grading. That small interface detail suggests a larger product opportunity: show how a quality decision changes the stock available to fulfil customer orders.

Topics: LLM Evaluation

Entities: LLM Evaluation

Mitchell Bryson AI Reliability Articles · September 05, 2026 00:00

The smartest systems are learning to game the test - Mitchell Bryson

One story formalised Fermat's Last Theorem, while others showed agents discussing escape, benchmarks disagreeing and an AI attack moving into spam. Together they point to a shift: when systems can optimise against the test, builders need verification with...

Topics: Benchmarks

Entities: Benchmarks