norgitov/ trends
Technology · people · ideas
RSS
RUEN
Back to discovery/Daily Papers3 hours ago
ResearchDaily Papers

S-Space: Exploring Spatial Workspace in Multimodal Language Models

The paper introduces S-Space, a linear subspace in multimodal language models where object-token representations encode continuous spatial coordinates. It demonstrates that S-Space serves as an internal workspace for spatial reasoning, enabling accurate spatial judgments through perception, internal reasoning, and selective verbal access, while highlighting limitations in spatial transformations across viewpoints.

Open original
SIGNAL FROM THE SOURCE
136
source votes
Tracking sinceOctober 12, 2026136 source votes
Momentum+40.44/hover 3.02 h
Discussion—Read comments ↗
PublishedOctober 9, 2026Zhiyuan Shi, Xiang Li, Xingzhuo Guo
BEHIND THE NUMBERS

How interest changes

Daily Papers

History starts here

The chart will appear after repeat observations. The current metric comes from the source.

136 source votes

Real observations only. History before source connection is not reconstructed.

WHY IT MAY MATTER

This research provides insights into how multimodal language models internally represent and process spatial information, which could inform the development of more robust spatial reasoning capabilities in AI systems.

A useful discovery?
KEEP EXPLORING

Connected ideas

Explore topic
Daily Papers

Continually Post-training Small Models via Synthetic Self-distillation

Continual post-training faces challenges of acquiring new capabilities while retaining prior knowledge. A method called Synthetic Self-distillation is introduced, providing high-quality supervision for small models without disrupting their original distribution. Experiments show the method outperforms six other approaches.

3 days agoDaily Papers

CheckerBench: Can Long-Horizon Agents Synthesize Static-Analysis Checkers?

CheckerBench is a benchmark for evaluating coding agents' ability to synthesize static-analysis checkers from defect specifications, including tasks derived from 300 CVEs across multiple repositories and language ecosystems. CheckerLab, an evaluation framework, measures diagnostic contrast, patch localization, and tool use, with results showing that current agents struggle to develop reliable checkers, achieving a maximum Pass@1 of 45.33%.

6 days agoDaily Papers

Tetris3D: 3D Scene Generation With Objects That Fit Together

Tetris3D is a generative framework for 3D scene reconstruction from a single image, ensuring objects are geometrically and physically coherent. It uses explicit conditioning on surrounding objects' geometry and physical relationships, along with a physics-based dataset called ComOb for training and evaluation.

5 days agoDaily Papers

Rethinking Cross-Tokenizer On-Policy Distillation: From Alignment Coverage to Supervision Reliability

The paper investigates on-policy distillation (OPD) in cross-tokenizer settings, focusing on alignment coverage and supervision reliability. It finds that strict 1:1 token alignment covers most student-generated tokens despite vocabulary mismatches, and that restricting reverse KL to a top-16 subset of shared vocabulary achieves comparable accuracy to full shared-vocabulary OPD, while adding span supervision reduces accuracy.

6 days agoDaily Papers

Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition

The study explores Adversarial Heuristic Learning (AHL), where AI agents refine game strategies with fixed model weights. The AAArena benchmark includes 12 games and 1,920 archived programs, evaluated through protocols mimicking real competitions. Results show agents can improve strategies but face challenges in complex games.

4 days agoDaily Papers

Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks

The paper introduces Memento 3, a system enabling frozen LLM agents to continuously learn explicit world models through external memory. The agent maintains a natural-language rulebook as persistent semantic memory, which is compiled into executable code for prediction and planning. Through a loop of observation, reflection, and verification, the agent refines its model and uses verified updates to guide interaction. On ARC-AGI-3, the single-model agent achieves 100.0 mean Relative Human Action Efficiency and uses 44% of the human action count. In an Atari Pong case study, a learned feedback controller wins 21:0 in each of three evaluated episodes.

4 days ago