SearcharxivSearch

arXiv subjects

Igor Ivanov

Publications and source records attributed to Igor Ivanov.

13 recordsLinked to original sources

LURE: Live-Usage Replay Evaluations for Reducing Evaluation Awareness

Large language models can recognize when they are being evaluated (evaluation awareness) and behave differently because of that, which undermines the validity of safety and alignment benchmarks. We propose LURE (Live-Usage Replay Evaluations), a method for constructing deployment-like evaluations by replaying realistic agentic interaction trajectories and appending evaluation prompt at the end. We also introduce an automated pipeline for measuring evaluation realism, combining detection of verbalized evaluation awareness and judge-model estimates of the probability of logs being an evaluation, and validate it on a large dataset of deployment and evaluation transcripts. We find that LURE-based evaluations are substantially less distinguishable from deployment than widely used benchmarks and synthetic evaluation generators, and can approach the realism of real conversations with users. We instantiate LURE in scheming, AI safety sabotage, and sycophancy settings. Our results suggest that evaluation realism is a crucial property of alignment benchmarks and should be reported alongside benchmark results, especially when such results are used in safety cases.

cs.CL

LLMs are Capable of Misaligned Behavior Under Explicit Prohibition and Surveillance

In this paper, LLMs are tasked with completing an impossible quiz, while they are in a sandbox, monitored, told about these measures and instructed not to cheat. Some frontier LLMs cheat consistently and attempt to circumvent restrictions despite everything. The results reveal a fundamental tension between goal-directed behavior and alignment in current LLMs. The code and evaluation logs are available at github.com/baceolus/cheating_evals

cs.AI

High-Precision Physics Experiments at Huizhou Large-Scale Scientific Facilities

In response to the capabilities presented by the High-Intensity Heavy Ion Accelerator Facility (HIAF) and the Accelerator-Driven Subcritical System (CiADS), as well as the proposed Chinese Advanced Nuclear Physics Research Facility (CNUF), we are assembling a consortium of experts in relevant discipline--both domestically and internationally--to delineate high-precision physics experiments that leverage the state-of-the-art research environment afforded by CNUF. Our focus encompasses six primary domains of inquiry: hadron physics--including endeavors such as the super eta factory and investigations into light hadron structures; muon physics; neutrino physics; neutron physics; the testing of fundamental symmetries; and the exploration of quantum effects within nuclear physics, along with the utilization of vortex accelerators. We aim to foster a well-rounded portfolio of large, medium, and small-scale projects, thus unlocking new scientific avenues and optimizing the potential of the Huizhou large scientific facility. The aspiration for international leadership in scientific research will be a guiding principle in our strategic planning. This initiative will serve as a foundational reference for the Institute of Modern Physics in its strategic planning and goal-setting, ensuring alignment with its developmental objectives while striving to secure a competitive edge in technological advancement. Our ambition is to engage in substantive research within these realms of high-precision physics, to pursue groundbreaking discoveries, and to stimulate progress in China's nuclear physics landscape, positioning Huizhou as a preeminent global hub for advanced nuclear physics research.

hep-ph

Resurrecting saturated LLM benchmarks with adversarial encoding

Recent work showed that small changes in benchmark questions can reduce LLMs' reasoning and recall. We explore two such changes: pairing questions and adding more answer options, on three benchmarks: WMDP-bio, GPQA, and MMLU variants. We find that for more capable models, these predictably reduce performance, essentially heightening the performance ceiling of a benchmark and unsaturating it again. We suggest this approach can resurrect old benchmarks.

cs.LG

Attosecond Delays in X-ray Molecular Ionization

The photoelectric effect is not truly instantaneous, but exhibits attosecond delays that can reveal complex molecular dynamics. Sub-femtosecond duration light pulses provide the requisite tools to resolve the dynamics of photoionization. Accordingly, the past decade has produced a large volume of work on photoionization delays following single photon absorption of an extreme ultraviolet (XUV) photon. However, the measurement of time-resolved core-level photoionization remained out of reach. The required x-ray photon energies needed for core-level photoionization were not available with attosecond tabletop sources. We have now measured the x-ray photoemission delay of core-level electrons, and here report unexpectedly large delays, ranging up to 700 attoseconds in NO near the oxygen K-shell threshold. These measurements exploit attosecond soft x-ray pulses from a free-electron laser (XFEL) to scan across the entire region near the K-shell threshold. Furthermore, we find the delay spectrum is richly modulated, suggesting several contributions including transient trapping of the photoelectron due to shape resonances, collisions with the Auger-Meitner electron that is emitted in the rapid non-radiative relaxation of the molecule, and multi-electron scattering effects. The results demonstrate how x-ray attosecond experiments, supported by comprehensive theoretical modelling, can unravel the complex correlated dynamics of core-level photoionization.

physics.atom-ph

Strong field ionisation of Argon: Electron momentum spectra and nondipole effects

We investigate the influence of relativistic nondipole effects on the photoelectron spectra of argon, particularly in the low kinetic energy region (0 eV - 5 eV). In our experiment, we use intense linearly polarised 800 nm laser pulse to ionise Ar from a jet and we record photoelectron energy and momentum distributions using a reaction microscope (REMI). Our measurements show that nondipole effect can cause an energy-dependent asymmetry along the laser propagation direction in the photoelectron energy and momentum spectra. Model simulation based on time-dependent Dirac equation (TDDE) can reproduce our measurement results. The electron trajectory analysis based on classical model reveals that the photoelectron which obtains negative momentum shift along laser propagation direction is caused by the interplay between the Lorenz force induced radiation pressure during its free propagation in continuum and re-scattering by Coulomb potential of the parent ion when it is driven back by the laser field.

physics.atom-ph

Delay time and Non-Adiabatic Calibration of the Attoclock. Multiphoton process versus tunneling in strong field interaction

The measurement of the tunneling time in attosecond experiments, termed attoclock, triggered a hot debate about the tunneling time, the role of time in quantum mechanics, where the interaction with the laser pulse involves two regimes of a different character, the multiphoton and the tunneling (field-) ionization. In the adiabatic field calibration, one of us (O. K.) developed in earlier works a real tunneling time model and showed that the model fits well to the experimental data of Landsmann et al. (Optica {\bf 1}, 343 2014). In the present work, it is shown that the model explains the experimental result in the nonadiabatic field calibration, where one reaches a good agreement with the experimental data of Hofmann et al. (J. of Mod. Opt. {\bf 66}, 1052, 2019). Furthermore, we confirm the result with the numerical integration of the time-dependent Schr\"odinger equation. The model is appealing because it offers a clear picture of the multiphoton and tunneling field-ionization regimes. In the nonadiabatic case (the nonadiabatic field calibration), the ionization is mainly driven by multiphoton absorption. Surprisingly, at a field strength $F \le F_a$ ($F_a$ is the atomic field strength) the model always predicts a time delay with respect to the quantum limit $\tau_a$ at $F=F_a$. For an adiabatic tunneling the saturation at the limit ($F=F_a$) explains the well-known Hartman effect or Hartman paradox.

quant-ph

Relativistic non-dipole effects in strong-field atomic ionization at moderate intensities

We present a detailed experimental and theoretical study on the relativistic non-dipole effects in strong-field atomic ionisation by near-infrared linearly-polarised few-cycle laser pulses in the intensity range 1014 -1015 W/cm2. We record high-resolution photoelectron momentum distributions of argon using a reaction microscope and compare our measurements with a truly ab-initio fully relativistic 3D model based on the time-dependent Dirac equation. We observe counter-intuitive peak shifts of the transverse electron momentum distribution in the direction opposite to that of laser propagation as a function of laser intensity and demonstrate an excellent agreement between experimental results and theoretical predictions.

physics.atom-ph

Attosecond angular streaking and tunnelling time in atomic hydrogen

Tunnelling, one of the key features of quantum mechanics, ignited an ongoing debate about the value, meaning and interpretation of 'tunnelling time'. Until recently the debate was purely theoretical, with the process considered to be instantaneous for all practical purposes. This changed with the development of ultrafast lasers and in particular, the 'attoclock' technique that is used to probe the attosecond dynamics of electrons. Although the initial attoclock measurements hinted at instantaneous tunnelling, later experiments contradicted those findings, claiming to have measured finite tunnelling times. In each case these measurements were performed with multi-electron atoms. Atomic hydrogen (H), the simplest atomic system with a single electron, can be 'exactly' (subject only to numerical limitations) modelled using numerical solutions of the 3D-TDSE with measured experimental parameters and acts as a convenient benchmark for both accurate experimental measurements and calculations. Here we report the first attoclock experiment performed on H and find that our experimentally determined offset angles are in excellent agreement with accurate 3D-TDSE simulations performed using our experimental pulse parameters. The same simulations with a short-range Yukawa potential result in zero offset angles for all intensities. We conclude that the offset angle measured in the attoclock experiments originates entirely from electron scattering by the long-range Coulomb potential with no contribution from tunnelling time delay. That conclusion is supported by empirical observation that the electron offset angles follow closely the simple formula for the deflection angle of electrons undergoing classical Rutherford scattering by the Coulomb potential. Thus we confirm that, in H, tunnelling is instantaneous (with an upperbound of 1.8 as) within our experimental and numerical uncertainty.

physics.atom-ph

Time delay anisotropy in photoelectron emission from the isotropic ground state of helium

Time delays of electrons emitted from an isotropic initial state and leaving behind an isotropic ion are assumed to be angle-independent. Using an interferometric method involving XUV attosecond pulse trains and an IR probe field in combination with a detection scheme, which allows for full 3D momentum resolution, we show that measured time delays between electrons liberated from the $1s^2$ spherically symmetric ground state of helium depend on the emission direction of the electrons relative to the linear polarization axis of the ionizing XUV light. Such time-delay anisotropy, for which we measure values as large as 60 attoseconds, is caused by the interplay between final quantum states with different symmetry and arises naturally whenever the photoionization process involves the exchange of more than one photon in the field of the parent-ion. With the support of accurate theoretical models, the angular dependence of the time delay is attributed to small phase differences that are induced in the laser-driven continuum transitions to the final states. Since most measurement techniques tracing attosecond electron dynamics involve the exchange of at least two photons, this is a general, significant, and initially unexpected effect that must be taken into account in all such photoionization measurements.

physics.atom-ph

Interpreting Attoclock Measurements of Tunnelling Times

Resolving in time the dynamics of light absorption by atoms and molecules, and the electronic rearrangement this induces, is among the most challenging goals of attosecond spectroscopy. The attoclock is an elegant approach to this problem, which encodes ionization times in the strong-field regime. However, the accurate reconstruction of these times from experimental data presents a formidable theoretical challenge. Here, we solve this problem by combining analytical theory with ab-initio numerical simulations. We apply our theory to numerical attoclock experiments on the hydrogen atom to extract ionization time delays and analyse their nature. Strong field ionization is often viewed as optical tunnelling through the barrier created by the field and the core potential. We show that, in the hydrogen atom, optical tunnelling is instantaneous. By calibrating the attoclock using the hydrogen atom, our method opens the way to identify possible delays associated with multielectron dynamics during strong-field ionization.

physics.atom-ph

Laser-sub-cycle two-dimensional electron momentum mapping using orthogonal two-color fields

The two-dimensional sub-cycle-time to electron momentum mapping provided by orthogonal two-color laser fields is applied to photoelectron spectroscopy. Using neon as the example we gain experimental access to the dynamics of emitted electron wave packets in electron momenta spectra measured by coincidence momentum imaging. We demonstrate the opportunities provided by this time-to-momentum mapping by investigating the influence of the parent ion on the emitted electrons on laser-sub-cycle times. It is found that depending on their sub-cycle birth time the trajectories of photoelectrons are affected differently by the ion's Coulomb field.

physics.atom-ph

Review of Strong Field Approximation and Investigating Semiclassical Evolution Approach

This paper theoretically analyzes the behavior of an atom driven by a strong electro-magnetic field. Moreover, besides traditional quantum mechanics method, we also investigate semiclassical approaches to this problem. We first performed strong field approximation for system of an atom driven by a strong electromagnetic field in velocity gauge. Our simulation result is consistent with theories and close to experiments except some reasonable difference caused by different parameters and omitted bound and final states in the transition amplitude. Next, a new semiclassical approach is used to solve Volkov wave function. We prove that semiclassical approximation works well in predicting the particle evolution in quantum world, especially for system in a strong electromagnetic field with low frequency. Finally, we also briefly illustrated how to use semiclassical approximation to get the same results as strong field approximation and also partly tested the viability of the semiclassical approach to Teller Potential. This part needs future work to accomplish. But still, semiclassical approximation provides a potentially new method to solve complicated system, which might be more effective than traditional quantum mechanics recipe.

physics.atom-ph