UofT Jinesis Lab to Present 3 Papers at COLM
ZhijingJin · x · 2026-07-17
The University of Toronto's Jinesis Lab announced that 3 papers will be presented at COLM 2026, covering the following primary research areas:
- AI Safety and Agent Defense: Proposed the ODILE method, defending against agent prompt injection attacks via tool call embeddings with orthogonal corruption injection.
- AI for Science: Released Stargazer, a scalable model fitting benchmark environment for evaluating AI agents under astrophysical constraints.
- LLM Safety Guardrails: Research found that LLM safety defenses can be bypassed token-by-token using the "incremental completion decomposition" technique.
More from Safety
- An MCP server signs every AI agent tool call into a verifiable Merkle chain — Funky_Chicken_22 · 2026-07-22
- AI industry astroturfing roundup tracks the sector’s fake-grassroots problem — ShakeelHashim · 2026-07-22
- New paper defines self-state attacks, showing OS defenses leave four agent-memory cases indistinguishable — Justgototheeffinmoon · 2026-07-22
- Substack starts labeling AI-generated or AI-influenced writing — StewartalsopIII · 2026-07-22
- ControlAI CEO says an international ban on superintelligence is needed to avert extinction risk — zetalyrae · 2026-07-22
- Coding agents are heading toward an AI-writes, AI-reviews, human-approves workflow — aftahi_ai · 2026-07-22