Never miss a research paper that matters to you. Paperzilla scans arXiv, bioRxiv, medRxiv, etc., daily, classifies must-read papers, and emails you a digest.
- Tweets189
- Following199
- Followers31
- Likes595
ALT Diagram of face-guided voice selection. The input shows overlapping blue and red speech waveforms. Selecting face A produces blue speech output; switching to face B produces red output, then selecting A returns to blue. Vertical markers align face selections with output changes.
ALT Screenshot of a paper titled “Can Open Be Too Open? On the Security and Trustworthiness of Fully Transparent LLMs,” by Sheila Zingg, Florian Tramèr, Martin Vechev, and Srdjan Čapkun of ETH Zurich. Below, two columns of text discuss transparency, reproducibility, trust, and security risks in LLMs.
ALT First page of “Is the arXiv Apocalypse upon us? A break in preprint growth coincident with agentic AI” by Carlo R. Contaldi, dated October 27, 2026. The abstract reports accelerated arXiv submission growth, largest in mathematics, and a shift toward single-author papers.
ALT Diagram titled “Online RL Training.” Real cloud-hosted iOS devices generate rollout trajectories evaluated by a “VLM-as-a-Judge.” Gemini-3.5-Flash provides dense rewards; Gemini-3.1-Pro provides terminal rewards. Both feed into PPO / GiGPO.
ALT Paperzilla is now officially in the Claude directory
ALT medRxiv page for “Decision-only medical AI: early evidence and pairing with language models,” showing authors, a DOI, and the selected Abstract tab. A prominent notice states that the preprint has not been peer-reviewed and should not guide clinical practice.
ALT In our modified CybORG CAGE 2 environment (Kiely et al., 2023), a red agent attacks network hosts while a blue agent defends. The blue agent’s RL policy (πRL) is trained to maximize reward but has no direct incentive to protect Confidentiality, Integrity, or Availability (CIA)-critical servers: Database (DS), Front-End (WS), and Auth (AS). To align an agent to these CIA operational priorities without retraining, we construct an alignment policy (πalign) using attribute predictors that score each action’s CIA impact, and interpolate it with the RL policy at test time using policy shaping.
ALT Self-improved red teaming based on role-sepration. The target session uses its live execution context to diagnose why an attack failed. The diagnosis is then transferred to a separate attacker that generates the next adversarial candidate. This preserves execution-grounded criticism while avoiding refusal or sandbagging that can arise when the target itself is asked to operationalize the attack.
ALT SoK: Design-Level Backdoors in Block Ciphers Constructions, Detection, and Key Recovery
ALT Correlation-1 differential-linear trail through the 16-round KG-DES, one row per round: the nibble activity of (∆Lr, ∆Rr) after r rounds, and the S-boxes active in round r + 1. Red: nibble difference possibly nonzero, and then always in {0, 2, 4, 6}; red S-boxes receive an input difference in U∗ . Blue: the mask λ = 80000000 on R16, which reaches L15 through the Feistel XOR and the round-16 S-box S4 as γ = 1 through P ⊤; that S-box, marked Em, carries both colours. Round keys, IP and the final swap are not drawn. From r = 5 on, every cell is red. RU = 15, RM = 1, RL = 0; p = r = q = 1.
ALT Diagram labeled “Figure 1: The round function Rᵣ of FRAST.” Inputs y₁, y₂, …, y₃₂ flow downward to updated outputs. Branches pass through round-key additions and S_erf blocks; a sum of branches feeds a round-key addition and S_crf block back into the first branch.
ALT Diagram of an attacker controlling third-party content accessed by an LLM agent. The agent invokes Skill A, which invokes Skill B, then Skill C. All three skills are marked “Clean,” but their composition leads to a “Violation.” A crossed-out path states “No Skill Modification.”
ALT Flowchart linking incrimination → detection → deterrence. Revealing malicious behavior and detecting unauthorized actions lead to gathering evidence that an agent is dangerous. Detecting and mitigating unauthorized actions and reducing attempted attacks lead to preventing agents from carrying out successful attacks.
ALT A case from SKILLCLASH. Top: two skills for the same job, with each rule shortened and labeled by its kind (Section 3). Bottom: outputs of Qwen3-4B for the request “Write the commit message: add pagination to the audit log API. Ticket AUD-77. Author: Omar Haddad. Component: audit-api.” Lines are shaded by the skill whose rule they follow (blue for A, orange for B, gray for the shared rule). Left: with both skills loaded, the model follows all six rules of B and only the shared rule of A, in either load order. Right: with one meta-skill written by Claude Opus 5.5 (Section 5), with A as the primary, the output follows all ten achievable rules.
ALT Five conceptual stages of SPRA construction. A shared unit census precedes relation planning and role-specific filling; assembly joins source-linked records and stores external citations separately. Validation checks profile conformance, relation endpoints, and source-link integrity.
