Liquid AI released DSpark draft model checkpoints for three LFM2.5 targets by adding a ~300M-parameter speculative-decoding drafter that proposes a 9-token block for the target to verify. It reports up to 3.18x faster decoding on an H100 while keeping greedy-decoded outputs identical and benchmark accuracy unchanged. The checkpoints require self-hosting with DSpark-enabled llama.cpp or SGLang, and speedups vary by acceptance rate (notably lower on MoE workloads).
AI systems solving advanced mathematics has triggered an existential crisis among some leading mathematicians about what mathematics is and what mathematicians will do next. The debate followed OpenAI’s publication of “10 Advances in Mathematics and Theoretical Computer Science.” Funding and employment models may shift as AI labs increasingly treat math proofs as computable tasks, while human training and grant value face new scrutiny.
Prof Jie Tang argued that parameter count alone is not enough to predict model capability and said GLM 5.3’s gains come mainly from RL on long-horizon, production-like environments rather than bigger models. He highlighted that advanced skills can require carrying causal chains of 20+ inference steps without losing the thread. The takeaway is that scaling focus shifts toward post-training data quality, effective compute, and environment design (including RL reward/verification), not just parameter size.
Cross-lingual knowledge transfer is presented as essential for training multilingual language models on low-data languages, because key knowledge for downstream tasks must come mainly from a high-resource language when target data is scarce. No specific number, date, or benchmark appears in the provided excerpt. The approach described shifts to using lexical interventions to move knowledge across languages under data constraints.
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.
