The paper investigates whether experience gained by large‑language‑model agents during molecular optimization in one protein pocket can be transferred to new, unseen pockets. By logging every edit and docking step, the authors build an experience base keyed on contact sites, the edit action, and the round, and find that this key sharply distinguishes improving from worsening edits. However, transfer is limited: only coarse, non‑specific descriptors of the edit and resulting molecule provide useful signals, while pocket‑specific information is largely irrelevant for predicting success.
arXiv:2606. 00555v1 Announce Type: new Abstract: Structure-based drug design increasingly employs LLM agents to iteratively refine ligands against a target pocket, yet a viable ligand must satisfy two often-conflicting objectives -- binding affinity and druggability -- which single optimization steps rarely improve together.
The paper introduces AgentDiff, a metric that quantifies how much LLM agents’ answers differ when inputs are altered by meaning‑bearing rewrites (paraphrases, synonym substitutions) versus presentation changes (reordering, formatting, distractors). Across 68 model–benchmark–scaffold combinations involving ten LLMs and over 1,500 questions, meaning‑bearing rewrites consistently produce a roughly 20‑percentage‑point higher inconsistency rate than presentation changes, a gap that persists across severity proxies and remains significant even outside the Qwen family. Trace analysis reveals that meaning‑bearing rewrites preserve the first action but reduce thought similarity from the second step onward, extending the divergence cascade—a phenomenon termed “stealth divergence.”
Budgeted Quotient-Residual Guidance (QRG) is an inference-time correction that enables quotient-based lead‑optimization objectives—such as distances, contacts, and anchored substructures—to be actively used with frozen pocket‑conditioned molecular diffusion models without retraining. QRG lifts quotient covectors to metric‑horizontal ambient directions and delivers them through a trust budget set by the sampler’s step norm, providing a closed‑form sampler‑budget update and a product‑budget split for section and residual controls. Experiments on frozen TargetDiff backbones show that QRG improves ligand‑generation quality on fragment growing, scaffold hopping, and linker design while preserving novelty and diversity.
arXiv:2606. 17107v1 Announce Type: cross Abstract: Prefix caching reuses prefill only across an exactly shared prefix, so one changed field invalidates the entire downstream cache.
arXiv:2610.15738v1 Announce Type: new Abstract: LLM optimization loops often show the proposer the candidates it has already tried and rejected, but leave it to the proposer to avoid them. We ask whe...
The study evaluates whether a portfolio of compact, semantically named descriptor blocks can match the performance of a 2048‑dimensional CheMeleon embedding in low‑data molecular assays. Using a fixed 11‑dimensional physicochemical base and greedily adding provenance‑screened blocks, the portfolio achieves a mean test AUC of 0.762 across nine ADME/Tox assays, comparable to CheMeleon’s 0.764 and better than Mordred’s 0.756. The results meet a predeclared pooled parity threshold but not all per‑assay thresholds, and further analysis confirms the competitiveness of the auditable representation while highlighting unresolved assay‑level differences.
The paper investigates how sequential knowledge editing can degrade a language model’s ability to discern reliable evidence from unreliable evidence without affecting overall accuracy. Using a conservatively tuned LoRA on Qwen2.5‑7B‑Instruct, the authors show that after 1,000 edits the model’s arbitration score for untouched facts drops by 36%, leading to higher error rates on its most confident decisions, while MMLU accuracy remains unchanged. The study also finds that in some model‑method combinations, sequential edits can reduce MMLU to chance levels even though edit success and locality remain perfect.
arXiv:2606. 20737v2 Announce Type: replace Abstract: We study factual edit propagation in a controlled synthetic knowledge-graph QA setting using a 2x2 grid that crosses loop recurrence with shared-memory access: a dense transformer (Dense), a looped transformer (Loop), a dense backbone with shared memory (Dense+Mem), and a looped backbone with shared memory (loop-memory coupling, LMC).
arXiv:2605. 07284v2 Announce Type: replace Abstract: A late-layer change learned during post-training may work on the base model's earlier state, or it may depend on earlier computation learned with it.
The paper introduces CELLAUDIT, a method for auditing whether inputs claimed to influence predictive models actually do so. By testing if an input can enter the computation, whether predictions depend on it, and if that dependence improves observed responses, the authors evaluate agent-generated predictors on a morphology‑transcriptomics benchmark (BBBC047). Their findings show that many models claim compound contributions that are not supported by the data, and that falsification‑guided revisions can recover genuine input effects while improving performance.
PMO-Dock introduces a standardized benchmark for docking‑based molecular optimization, comprising 25 tasks that span hit generation, lead optimization, and a specificity task that rewards strong on‑target binding while penalizing off‑target interactions. The protocol separates model development from final evaluation, using validation tasks for hyperparameter tuning and held‑out test tasks evaluated under a strict oracle budget to mimic realistic drug‑discovery constraints. Four diverse methods—Saturn, GenMol, Genetic‑guided GFlowNet, and ChemLlama—were benchmarked, revealing no single method dominates across all tasks and highlighting limited hyperparameter transferability, which underscores the need for method‑dependent selection strategies.
The paper investigates whether the locate‑then‑edit approach for knowledge editing, previously applied only to autoregressive language models, can be transferred to masked diffusion models (MDMs). It finds that the optimal edit location—an early‑to‑mid‑layer MLP at the last subject token—remains the same for both model types, but that MDMs suffer a sharper decline in performance when editing longer, multi‑token facts. By incorporating intermediate partially‑unmasked states into the edit optimization, the authors restore multi‑token editing performance in MDMs.