$ cat briefs/daily/2026-11-10.md
2026-11-10
November 10, 2026 (Tue)
1 top story · 1 paper pick · 1 evaluation signal
**Reconstructed archive:** prepared on **2026-11-15 (KST)** after the original automation failed, using the original morning cutoff of **2026-11-10T08:00:00+09:00** (**2026-11-09 23:00 UTC**). This issue was not published or emailed on the missed date; no subscriber delivery is authorized. **Provenance and limits:** the research below is preserved in the **2026-11-09** Hugging Face snapshot. Its bytes match repository commit `41eadb43ca1714722dcd5275cf75340f7ecf8ae7`, timestamped **2026-11-08 22:53:16 UTC**, before the cutoff. The version-specific primary records linked below confirm earlier submissions; these are not releases on the issue date. Coverage is limited to abstracts and metadata, without independent verification of full papers, code or results. This selective reconstruction cannot recover the complete morning feed. Items without verifiable pre-cutoff availability are omitted. Wiki links provide retrospective topic context and may contain later updates. ([snapshot](../../sources/papers-daily/hf-daily-2026-11-09.md))
Top Stories
1. DPDM separates language representations from decisions based on experience
A frozen LimiX-2 predictor uses evidence represented by Qwen3.5-27B without updating either model's parameters during deployment. Across 50 benchmarks, authors report 78.80% mean accuracy with few-shot prompting versus 84.92% for DPDM at their respective history budgets; DPDM reaches 82.33% with identical histories and queries.
Why it matters: the matched-history result helps distinguish prediction capability from access to more evidence. These are author-reported comparisons, not proof of reliable deployment. Version history: v1, 2026-11-01 17:27:21 UTC. → Agents (LLM Agents) (snapshot) (primary v1)
Paper Picks
SLA: align a pi05-based robot policy's visual embeddings with a frozen 3D perception model during continual pre-training. Auxiliary components are used only during training, preserving the original inference architecture. Why read it: spatial supervision can change training without adding those components to deployment; the abstract supplies no numerical performance results. Version history: v1, 2026-11-05 17:52:16 UTC. → Embodied Agents (snapshot) (primary v1)
Watch
- VisSearch-Bench: authors report that the best of 16 evaluated multimodal models fully solves 26.47% of 204 visual-search tasks. Complete-task accuracy is distinct from their score for intermediate results. Why it matters: partial progress should not be mistaken for a complete answer. Version history: v1, 2026-11-03 13:40:41 UTC. → Eval Harness Configuration (snapshot) (primary v1)
New in Wiki
- Original dated work: the retained log records the creation of the Harness-S and Hybrid Computer Use paper pages on 2026-11-08, not 2026-11-10. Reconstruction: no wiki pages were created; this issue links existing topic pages. (log)
Updates
- Original dated work: the 2026-11-08 ingest entry records updates to Agents (LLM Agents) and Eval Harness Configuration; no 2026-11-10 ingest entry is retained. Reconstruction: no wiki pages were edited. An absent entry does not establish an absence of news. (log)