generational-genetic-search-over-llm-agent-generated-scientific-hypotheses-drug-repurposing CLM-0398observationsingle paperReviewed by AI
In a generational genetic search over LLM-agent-generated scientific hypotheses (drug repurposing across 34 cancer types), replacing fitness-guided parent selection with uniform random selection while holding the scientific roles, variation operators and hypothesis count fixed lowers the population's mean and minimum external evidence scores but leaves the maximum unchanged, and removing crossover or mutation altogether produces mixed shifts that do not survive multiple-testing correction — so the selection rule, not the semantic variation operators, is where the measurable coordination gain sits.
Capability: Getting more from several models than from one
Observed on
One base model (gpt-5.4-mini), population 6, 3 generations, drug repurposing hypotheses scored post hoc by DepMap selectivity and Open Targets association; operator ablations on on.
Sources
- Validated by: Mean DepMap selectivity 0.171 and Open Targets association 0.426 versus 0.115/0.329 for the strongest of six task-matched baselines (Tree of Thoughts) and 0.039/0.163 for single-pass generation, on 26 and 29 shared cancer types with the selected hypothesis chosen before external scoring; gains repeated on 27 held-out cancer types.. Measured: 94 runs, 34 cancer types, two post-hoc biological metrics never used as search fitness, six shared-protocol baselines. The selection ablation is the only clean single-variable comparison; [truncated]
Either way, bring evidence: a link, or what you saw.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- For a frozen-backbone multi-agent scientific discovery pipeline on HypoBench tasks, distilling reusable procedural skills separately per stage — from discriminator critique of a whole hypothesis batch for generation, and from execution outcomes compared to ground-truth results for testing — beats no skill, an LLM-written skill, and a hand-written skill, and removing each stage's feedback signal costs most of the gain.Improving its own scaffolding from failures · reviewed by AI
- When injecting revised medical guideline facts into 4B-14B instruct models by LoRA SFT under a matched token budget, rendering the same update events as extended-matching items — several related vignettes sharing one candidate answer pool — transfers better to a held-out temporally anchored oncology staging benchmark than short-answer, fill-in-the-blank, or per-vignette multiple-select renderings of the identical content.Absorbing and flexibly using new knowledge · reviewed by AI
- For selecting among multiple generated SQL candidates, a listwise selector given retrieved structured selection criteria distilled from training question-SQL pairs, with candidates grouped by execution result and permuted only within groups, reaches higher execution accuracy than pointwise, pairwise and multi-selector baselines on the same candidate pools while using far fewer selector calls and tokens.Writing a reusable locator: regex, SQL, XPath · reviewed by AI
- When automatically rewriting the prompts of a multi-agent LLM system from its failures, locating the responsible agent by injecting a hint into one agent at a time and using that agent's hint-corrected intermediate output as the supervision signal produces prompt updates that generalize better than critique derived only from the system-level output.Improving its own scaffolding from failures · reviewed by AI
- When medical LLMs are given a clinical practice guideline in context and asked to produce the decision pathway a case requires, final-outcome accuracy does not track adherence to the guideline-defined step sequence: models can reach the right outcome while omitting, reordering, skipping or padding steps, and medically fine-tuned models score no better on pathway consistency than the general models they were tuned from.Competence in a specialist domain · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: An experiment where random parent selection matches fitness-guided selection on mean and minimum external scores, or where removing crossover and mutation causes a consistent significant drop on both metrics across a larger set of cancer types.