Tools for this
Browse all ->Fresh stories
Vik Paruchuri challenges ParseBench-2’s OCR scoring methodology
Vik Paruchuri argues that ParseBench-2’s page-number requirements reward parser defaults rather than OCR quality. The creators defend the open methodology while acknowledging annotation and metric issues and updating the leaderboard.

Sentence Transformers v6.3 adds pooling-aware training for retrieval with fewer vectors
Sentence Transformers v6.3 adds pooling-aware training for multi-vector retrieval. Maintainer tests retained 97.3% of unpooled retrieval quality with 84.7% fewer document vectors; code and checkpoints are available.

Artificial Analysis updates Decision Index v2 with confidence-aware scoring
Artificial Analysis's Decision Index v2 adds confidence-aware scoring and raises private-dataset coverage to 45%. Its AA-Paperclip test requires all 15 answers for a document to be correct.


EmbeddingGemma 2 maps text, code, images, video and audio into one embedding space
A technical breakdown describes EmbeddingGemma 2’s shared embeddings for text, code, images, video and audio. It details staged distillation from Gemini Embedding 2, Matryoshka dimensions and int4 training.

Vik Paruchuri challenges ParseBench-2’s OCR scoring methodology
Vik Paruchuri argues that ParseBench-2’s page-number requirements reward parser defaults rather than OCR quality. The creators defend the open methodology while acknowledging annotation and metric issues and updating the leaderboard.

Perplexity releases Decider v1.3 with Locate text-span extraction
Perplexity released open weights for Decider v1.3 and its Flash variant. The new Locate feature returns matching text with character offsets for tasks such as PII detection, and Flash can run locally.

Rivet launches OpenOrange to run Cloudflare Workers on Node.js or Bun
Rivet's OpenOrange runs Cloudflare Workers on Node.js or Bun with Wrangler-compatible deployment and portable bindings. Rivet also provides live migration tooling and publishes workload cost comparisons against Cloudflare.
Sentence Transformers v6.3 adds pooling-aware training for retrieval with fewer vectors
Anthropic's Mythos 5.2 system card discloses deliberate limits on AI R&D capabilities
Practitioners turn agent feedback into reusable skill documents and lints
Artificial Analysis updates Decision Index v2 with confidence-aware scoring
Briefs forNovember 14

Daily AI Digest
Get the best stories delivered
to your inbox



