SCOPE
Self-play for open-ended tasks
Cyrus Kwan et al.
Learning without
curated answers
Inside text + image
attention heads
Composing multiple
behaviours
Beyond headline
benchmark scores
Interfaces for relational
generalisation
Self-play for open-ended tasks
Cyrus Kwan et al.
What retrieval heads actually write
Aryo Gema et al.
When benchmarks admit shortcuts
Matteo Attimonelli et al.
Grounding behavioural evaluation
Jonas Leo Mueller, Dario Zanca et al.
Can retrieval heads see images?
Aaron Branson Cigres Li et al.
Same answer, different representations
Farooq Ahmad Wani et al.
Controlling behaviours without interference
Giwon Hong et al.
What controls relational generalisation?
Alessandro Cagiano et al.
Self-play search distillation
Lorenzo Molfetta et al.
Chain-of-thought faithfulness
Aryo Gema et al.
Why verbose prompts help robustness
Farooq Ahmad Wani et al.
Cyrus Kwan et al.
Aryo Gema et al.
Matteo Attimonelli et al.
Jonas Leo Mueller, Dario Zanca et al.
Aaron Branson Cigres Li et al.
Farooq Ahmad Wani et al.
Giwon Hong et al.
Alessandro Cagiano et al.
Self-play for open-ended tasks
Controlling behaviours without interference
What controls relational generalisation?
What retrieval heads actually write
Can retrieval heads see images?
When benchmarks admit shortcuts
Grounding behavioural evaluation
Same answer, different representations
Cyrus Kwan et al.
Aryo Gema et al.
Matteo Attimonelli et al.
Jonas Leo Mueller, Dario Zanca et al.
Aaron Branson Cigres Li et al.
Farooq Ahmad Wani et al.
Giwon Hong et al.
Alessandro Cagiano et al.
SCOPE / Cyrus Kwan et al.
SCOPE
LOCOS
Composed image retrieval
AI psychophysics
Multimodal retrieval heads
Hidden VLM instability
Composing steering vectors
Graph-to-table interfaces
NeurIPS / EMNLP / LoG 2026
SPSD / FACE-Eval
VLM frequency filtering
SCOPE
LOCOS
Composed image retrieval
AI psychophysics
Multimodal retrieval heads
Hidden VLM instability
Composing steering vectors
Graph-to-table interfaces
Cyrus Kwan et al.
Aryo Gema et al.
Matteo Attimonelli et al.
Jonas Leo Mueller, Dario Zanca et al.
Aaron Branson Cigres Li et al.
Farooq Ahmad Wani et al.
Giwon Hong et al.
Alessandro Cagiano et al.
Self-play for open-ended tasks
Cyrus Kwan et al.
What retrieval heads actually write
Aryo Gema et al.
When benchmarks admit shortcuts
Matteo Attimonelli et al.
Grounding behavioural evaluation
Jonas Leo Mueller, Dario Zanca et al.
Can retrieval heads see images?
Aaron Branson Cigres Li et al.
Same answer, different representations
Farooq Ahmad Wani et al.
Controlling behaviours without interference
Giwon Hong et al.
What controls relational generalisation?
Alessandro Cagiano et al.
Cyrus Kwan et al.
NeurIPSAryo Gema et al.
NeurIPSMatteo Attimonelli et al.
NeurIPSJonas Leo Mueller, Dario Zanca et al.
NeurIPSAaron Branson Cigres Li et al.
EMNLPFarooq Ahmad Wani et al.
EMNLPGiwon Hong et al.
EMNLPAlessandro Cagiano et al.
LoGSCOPE
LOCOS
Composed image retrieval
AI psychophysics
Multimodal retrieval heads
Hidden VLM instability
Composing steering vectors
Graph-to-table interfaces
SCOPE
LOCOS
Composed image retrieval
AI psychophysics
Multimodal retrieval heads
Hidden VLM instability
Composing steering vectors
Graph-to-table interfaces
Self-play, retrieval, steering, evaluation,
and relational generalisation.