Chain-of-thought prompting degrades instruction-following accuracy on IFEval and ComplexBench across most tested models, and selective reasoning can partially recover the loss.
Metrology of weak quantum perturbations
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
We consider quantum systems with a Hamiltonian containing a weak perturbation i.e. $\boldsymbol{H=H_0} + \boldsymbol{\lambda} \cdot \boldsymbol{\tilde{H}}$, $\boldsymbol{\lambda}= \{\lambda_1, \lambda_2,...\}$, $\boldsymbol{\tilde{H}}$ $= \{H_1, H_2,...\}$, $\left|\boldsymbol{\lambda}\right| \ll 1$, and address situations where $\boldsymbol{\tilde{H}}$ is known but the values of the couplings $\boldsymbol{\lambda}$ are unknown, and should be determined by performing measurements on the system. We consider two scenarios: in the first one we assume that measurements are performed on a given stationary state of the system, e.g., the ground state, whereas in the second one an initial state is prepared and then measured after evolution. In both cases, we look for the optimal measurements to estimate the couplings and evaluate the ultimate limits to precision. In particular, we derive general results for one and two couplings, and analyze in details some specific qubit models. Our results indicates that dynamical estimation schemes may provide enhanced precision upon a suitable choice of the initial preparation and the interaction time.
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
When Thinking Fails: The Pitfalls of Reasoning for Instruction-Following in LLMs
Chain-of-thought prompting degrades instruction-following accuracy on IFEval and ComplexBench across most tested models, and selective reasoning can partially recover the loss.