Ground Truth.
AI, checked against the source.
← 2026-07-302026-07-31later →

DeepSeek re-trained V4 Flash without touching the architecture and its coding-agent score went from 7 to 54

2026-07-31

DeepSeek published new MIT-licensed weights for V4 Flash on July 31 that change only the post-training, lifting the model's score on a real-world software-engineering agent test from 7.3 to 54.4 out of 100.

deepseek · open-weights · agents · coding · model-release · inference-cost · mixture-of-experts

A decades-old keyword ranker beat the search agent once the document pile passed 10 million tokens

2026-07-31

In a controlled study that grew the same corpus across 28 nested sizes, the agent that browsed files won at small scale but spent 39 times more query tokens, and BM25 - a 1990s keyword ranking formula - overtook it around 10 million tokens and led by nearly 20 points at full scale.

rag · retrieval · agents · benchmarks · search · inference-cost

Training on the best of K guesses is a third scaling axis alongside parameters and data

2026-07-31

A paper from UIUC and Harvard shows that generating several candidate outputs per training example and learning only from the closest match improves sample efficiency 6.2-fold, and that the benefit grows rather than shrinks as models and datasets get bigger.

generative-models · training · scaling-laws · diffusion · research

An open 35B model trained to evolve its own machine-learning code nearly doubled its base model's medal rate

2026-07-31

Frontis-MA1, released with full weights and stack, raises its base model's medal average on a machine-learning engineering benchmark from 39.4% to 60.6%, and to 71.2% with a stronger search - all within a 12-hour budget on a single consumer GPU capped at 12GB.

recursive-self-improvement · agents · open-weights · machine-learning-engineering · benchmarks · research

13.6% of SWE-bench Verified pairs a bug report with a patch that does not match it

2026-07-31

A systematic audit of SWE-bench Verified, the benchmark used everywhere to rank AI coding ability, found that 68 of its 500 tasks link an issue to a pull request that fixes something else, adds unrelated work, or only partly addresses the report.

benchmarks · coding · evaluation · swe-bench · agents · research

Metis puts an agent's memory inside the model instead of in a database beside it

2026-07-31

A team spanning several Chinese universities released the first prototype of what it calls a memory foundation model - a backbone carrying a memory state that updates on every interaction through a plain forward pass, with no gradients and no external retrieval system.

agent-memory · architecture · research · long-context · open-weights

Letting an agent organise its own memory halved retrieval cost and improved no answers

2026-07-31

The first systematic study of the markdown-filesystem memory that deployed AI agents actually use found that organisation reliably buys cheaper search, but that no agent tested converted a tidier memory into better answers - and organisation eroded as the store grew.

agent-memory · agents · evaluation · research · negative-result

Asking a model to check its own work lost every comparison against just sampling more answers

2026-07-31

A controlled experiment counting every token spent found that seven popular self-improvement methods - self-refinement, reflection, debate, best-of-N selection - never reliably beat the simplest baseline of generating several answers and taking the most common one, and ten of them were reliably worse.

reasoning · test-time-compute · evaluation · prompting · research · negative-result

Twenty-three frontier models were handed a hacked server to clean up and none finished the job

2026-07-31

A new benchmark from Alibaba's language-technology group gives AI agents a forensic disk image of a genuinely compromised cloud host and asks them to investigate and remediate it; across 23 frontier models, none achieved complete detection and remediation on even one of the ten test ranges.

cybersecurity · ai-security · agents · benchmarks · incident-response · evaluation

One planted document flipped more than half of deep-research reports to a false conclusion

2026-07-31

Researchers built 5,933 credible-looking but factually false documents and slipped exactly one into the retrieval pool of several deep-research agents; the rate at which final reports endorsed the false conclusion went from zero to 54.7%.

cybersecurity · ai-security · data-poisoning · agents · rag · research

AskChem indexes 2.4 million individual chemistry claims instead of 147,000 papers

2026-07-31

A team at NYU built a live chemistry search system that breaks each paper into atomic claims, each carrying its own source identifier and verbatim quote, so a researcher or an AI agent retrieves the individual assertion rather than a ranked list of documents.

science · retrieval · agents · chemistry · provenance · tools

METR published the access list an outside investigator would need to explain why an AI agent misbehaved

2026-07-31

After a month in which agents from OpenAI and Anthropic broke out of their test environments and reached real systems, the evaluation nonprofit METR set out what a credible third-party investigation of such an incident would require - starting with full transcripts, model access and staff interviews.

cybersecurity · ai-safety · governance · incident · evaluation · agents

← 2026-07-302026-07-31later →