IBM released a technical walkthrough describing how its Granite 4.2 reasoning model family was built and trained. The first stage of long-context training extends the model’s context window to 512K tokens. The work shifts Granite from instruction-following toward explicit reasoning with dense decoder-only models that support thinking/non-thinking modes and native tool calling, and it adds multi-stage RL including agentic tool use for the 8B and 30B sizes.
Ropedia launched HOMIE Gen2, a head-mounted wearable that records egocentric human movement data to train robotic AI models. The device synchronizes four camera streams, spatial audio, and inertial data to within 50 microseconds. HOMIE Gen2 enables faster and cheaper deployment than the previous generation, with deployment said to be 10 times faster and costing about one-12th the price of a full lab capture.
Quantization-Aware Healing (QAH) showed that a GPT-OSS 120B model compressed to 60B parameters and quantized to MXFP4 can outperform its own recovered bfloat16 version when healed using KL distillation from the original pre-compression teacher instead of from the recovered checkpoint. It beats the 60B bfloat16 checkpoint on 7 of 9 benchmarks, including +7.4 on AA-LCR long-context reasoning and +5.6 on AIME 2025 math. This changes deployment pipelines by reducing reliance on quantization-aware training’s long, potentially unstable fine-tuning and enabling a safer, faster 4-bit recovery that also improves accuracy rather than only restoring it.
A team at Apple ran a large controlled distillation study and published Distillation Scaling Laws to model how student loss depends on teacher strength, student size, and training data.
Generalist AI Inc. reportedly raised $200 million to develop artificial intelligence software for robots. The reported funding round was led by 8VC, and it followed a previous $400 million round in June. Generalist’s new capital is expected to support Gen-1.5, its robotic-arm model that can be taught by demonstrations and simulated clips, reducing the time and code updates needed to create automation workflows.
Every AI story that matters,
in your inbox by 8am.
TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the
day in two minutes. Follow companies and topics for alerts, or get the
briefing in Slack. Free, no spam, unsubscribe anytime.
Reading TLDRocket needs no cookies, and the readership counts we rely on come from
our own cookieless analytics. Google Analytics is the exception: it sets cookies and
reports to Google, so it stays switched off until you allow it. You can change your
mind any time from “Cookie settings” in the footer.
Strictly necessary
Session security and form protection (tldrocket-session,
XSRF-TOKEN, 2 hours). The site cannot work without them,
so they need no consent.
Always on
Google Analytics 4 (_ga,
_ga_<id>, up to 2 years). Measures which
stories and sections readers use. Google acts as a third-party processor and may
store the data outside the EU. No advertising, no profiling, no data sold.
