Matching works

4 works

cs.AI — Artificial Intelligence

Automated Conjecturing: Generation, Selection and Evaluation

Contributed by Ayush Agrawal, Vedant Shah, Navin Goyal, Ankit Anand, Aaron Courville

Deciding which conjectures are worth proving is a crucial part of mathematical discovery, and a system that learns mathematics from its axioms alone must make that decision without human guidance. Self-play provers such as Minimo learn a conjecturer alongside the prover and reward it for producing hard conjectures, but a conjecture that is harder to prove is not necessarily more useful for improving a prover: the learned conjecturer drifts toward long, artificial statements. We study conjecturing in a controlled, single-round setting in the Peano formal environment, across propositional logic, abelian groups, and natural-number multiplication, evaluating a set of conjectures by how much a fixed prover improves on held-out theorems after training on their proofs. We show that a far simpler generator does better: making small structural edits to statements the prover already knows yields conjectures that, at an equal training budget and under best-first search, improve the prover by 6 to 13 percentage points more than Minimo's learned conjectures. The gain holds even when conjectures are drawn at random from the generated pool. In contrast to the common practice of selecting the most difficult statements, we also propose Hard Neighbours (HN), a score that favors conjectures whose proof difficulty grows disproportionately relative to their structural distance from easier reference statements; HN gives the largest selection gains in propositional logic, but no selection method is best in every domain. In the current setting, how conjectures are generated matters more than how they are selected, and structural perturbation is the baseline that future conjecturers should be measured against. Apart from improvements in proving ability, we also study other attributes of generated proofs: the amount of search required and the length of the proofs.

Primarily human-written textHuman understanding: all partsconjecturinglanguage models
Oct 29, 202640 pages
PDFView work

cs.AI — Artificial Intelligence

Do Self-Improving Coding Agents Beat Textbook Baselines? Evidence from a Darwin Gödel Machine Port and Two Published Systems

Contributed by Tejasvi Singh Tomar

Self-improving agent harnesses such as the Darwin Gödel Machine (DGM) let a coding agent edit its own scaffolding, and they keep the edits that score better and report large gains on coding benchmarks. In three pre-registered rounds, we compare what such loops find with standard engineering changes ("textbook moves"), one-shot redesigns by a language model, and placebo loops that make no edits; we also re-run released agents of the DGM and of ADAS with newer models. On hard LiveCodeBench problems with Claude Haiku 5.5, the starting agent often stops on an empty turn because its output-token limit is too small for the model's reasoning. A single textbook move, which raises this limit to 16,000 tokens and makes three other changes to the model loop, recovers about 90% of the gain of the best loop agent and costs less than half as much per task. Four of five one-shot redesigns by Claude Opus 5.5 also raise the limit, to 8,192 tokens, without evaluation feedback. The best loop agent still leads the textbook moves by 1–5 percentage points in pass rate. On held-out SWE-bench Verified tasks, run in the DGM's harness, the DGM's released best agent does not outperform its initial agent; three textbook moves on the initial agent outperform the best agent at a quarter of its cost, whether or not the patches of runs that ended early, most of them the best agent's, are kept. Claims of self-improvement should be measured against textbook moves, one-shot redesigns, placebo loops, held-out tasks and newer models.

A mix of human-written and AI-generated textHuman understanding: all parts
Oct 28, 202628 pages
PDFView work

cs.AI — Artificial Intelligence

BLLM: Beyond the LLM with Deep Human-Mind-AI

Contributed by Deep Bhattacharjee, Ushashi Bhattacharya

A system that improves itself keeps every change that passes a check, so a check that makes errors lets mistakes pile up. We compute the least checking that keeps such a system sound. It is an entropy of the costs of the changes, and an endless chain of improvements stays sound at finite cost exactly when that entropy is finite. A stack of checked layers costs more than its layers would cost alone. We also compute how many guesses a proposer needs before a right one appears: human authors come close to this least number, and language models need 1.5 to 4.3 times as many. Anaxia, an engine built on these laws, rebuilt the published 2016 counterexample to Steinberg's conjecture from the rules of a colouring problem alone. The engine is a work in progress, and its rules and search are being revised and extended for the open problems it has not yet settled.

A mix of human-written and AI-generated textHuman understanding: some partsautomated discoveryentropyrecursive self-improvementverification
Oct 20, 202675 pages
PDFView work

math.OC — Optimization and Control

A Finite Coordinate Reduction for Approachability Loss in Lean

Contributed by Arthur, David Barros Hulak, Ruy J. G. B. de Queiroz

We explain a Lean 4 development of a finite-coordinate loss-preserving construction motivated by Theorem 4 of Dann, Mansour, Mohri, Schneider, and Sivan. Nonempty finite constraint and action index sets give a joint probability simplex with a canonical action marginal. For each mixed constraint, a linear comparator adds its outer product with that marginal. The comparator has mass two and leaves the legal action set. A one-step pairing identity yields exact equality of finite-horizon approachability and comparator-regret objectives, with explicit anchored lifts and marginal decoders. The development also proves a row-wise decomposition with at most one rank-one term per constraint index. We describe the twelve declarations registered in Palomar and their historical verification provenance. The scope is deliberately narrow: the constraint simplex parametrizes coordinate functions, both causal strategy translations observe original loss histories, and the comparators have no fixed point in the joint simplex. Thus the checked statements establish finite algebraic loss preservation; they do not certify the full published reduction, its fixed-point-defined improper class, reduced-loss-only feedback, or asymptotic rate theory. No mathematical novelty or formalization priority is claimed.

Primarily AI-generated textHuman understanding: some parts
Oct 01, 20269 pages
PDFView work

Advanced search

Text
Human understanding
Linked formalizations