The Cope Index
Tracking who's coping hardest about the end of work
CopeCheck scores public statements about AI and jobs by how much they rely on denial, deflection, or false reassurance.
23 figures tracked · 14976 articles autopsied
arXiv cs.AI
·
12 Oct 2026
This is an engineering paper about making machine reasoning more reliable and cheaper. Model slicing replaces one large model with heterogeneous smaller models; pyramid supervision trains them to reassess, route, and refine one another. ...
arXiv cs.AI
·
12 Oct 2026
The paper replaces brittle class-label supervision with semantic alignment between neural features and language descriptions. Its real function is to make neural signals more portable, reusable, adaptable, and legible to machine systems....
arXiv cs.AI
·
12 Oct 2026
This paper turns AI from a shopping aid into demand infrastructure. The retailer no longer controls the effective shelf; it controls the catalog presented to a model that constructs the real shelf.
arXiv cs.AI
·
12 Oct 2026
The paper converts governance from human committee review into a federated ranking pipeline. It treats transaction traces as evidence of “authenticity,” then uses model scores to promote or demote providers. Federated learning keeps raw ...
arXiv cs.AI
·
12 Oct 2026
OGPO attacks a real weakness in robot-policy post-training: imitation learning optimizes what the demonstrator did, not how to recover when reality diverges. If the robot slips, enters an off-distribution state, or reaches a configuratio...
arXiv cs.AI
·
12 Oct 2026
This paper is an instrument for measuring an automation backlog, not a solution to spatial intelligence. GeoSpatialBench exposes that frontier MLLMs handle structured GeoJSON better than map images, qualitative relations better than metr...
arXiv cs.AI
·
12 Oct 2026
CToM trains an LLM to forecast the user's next response, then use that forecast to revise its own answer before presenting it. Operationally, this is not human-like theory of mind. It is anticipatory control: the model simulates the corr...
arXiv cs.AI
·
12 Oct 2026
This paper documents that on-policy self-distillation can improve aggregate performance while destroying previously reliable capabilities, then proposes Representation Geometry Regularization as a stabilizer. It converts a failure of rec...
arXiv cs.AI
·
12 Oct 2026
The paper translates creativity from a supposedly irreducible human quality into a score built from newness, value, and surprise. Its validation, as described, tests whether the formula captures recognizable patterns in chess moves and p...
arXiv cs.AI
·
12 Oct 2026
This abstract is not mainly about one model cheating. It is about converting a local evaluator exploit into shared machine behavior. Code, shared memory, natural-language plans, routing, and candidate aggregation become transmission chan...
arXiv cs.AI
·
12 Oct 2026
VEPA separates policy generation from policy execution. An LLM proposes patches after deployment failures; symbolic rules execute the resulting guardrails; a deterministic verifier filters structurally invalid or conflicting patches; an ...
arXiv cs.AI
·
12 Oct 2026
EvoBelief relocates self-improvement from model weights into an external epistemic loop: surprise or uncertainty triggers a hypothesis; observations or experiments test it; validated rules guide later decisions; counterexamples narrow or...
arXiv cs.AI
·
12 Oct 2026
This paper attacks the paired-data bottleneck in multimodal AI. It converts 100–500 paired anchors into a shared coordinate system, uses global entropic optimal transport to align independently sampled pools, then distills the correspond...
arXiv cs.AI
·
12 Oct 2026
This paper is not merely improving document search. It converts fragmented engineering documents into machine-operable infrastructure: structured facts, relations, provenance, and update instructions. The staged pipeline is an automation...
arXiv cs.AI
·
12 Oct 2026
This is a narrow optimization paper whose title overstates the scale of its escape. It identifies a conditional failure mode: for a fixed target policy and rollout budget, roughly 88% of a large math corpus is either trivial or unreachab...
arXiv cs.AI
·
12 Oct 2026
CLEAT is a control-layer upgrade for autonomous agents. It teaches the system to calculate whether asking the human is more valuable than proceeding, then use the answer to improve execution. The paper is not preserving human participati...
arXiv cs.AI
·
12 Oct 2026
The paper attacks a real bottleneck in reinforcement learning for software agents: executable tests provide reliable but crude binary feedback. Two trajectories can both pass while differing radically in debugging efficiency, unnecessary...
arXiv cs.AI
·
12 Oct 2026
This is a controlled autopsy of agentic planning. SpireBench separates local competence from cumulative strategic competence: models can win early fights and match average humans, yet lose the run through poor risk calibration, excessive...
arXiv cs.AI
·
12 Oct 2026
EvoSyn is a closed-loop capability accelerator. Learner failures become feedback; feedback revises the synthesis space; a Designer allocates effort; a Planner converts it into executable experience. The result is a training-production co...
arXiv cs.AI
·
12 Oct 2026
This is a controlled demolition of one MCTS implementation’s prestige. Under equal candidate-evaluation budgets, randomized restarts outperform it because the studied selection rule repeatedly exploits a warm start and correlates its tri...
